Network Optimization:

Continuous and Discrete Models

Dimitri P. Bertsekas
Massachusetts Institute of Technology

WWW site for book information and orders

http://www.athenasc.com

Athena Scientific, Belmont, Massachusetts

Athena Scientific
Post Office Box 391
Belmont, Mass. 02178-9998
U.S.A.

Email: [email protected]
WWW: http://www.athenasc.com

Cover Design: Ann Gallager

c 1998 Dimitri P. Bertsekas

All rights reserved. No part of this book may be reproduced in any form
by any electronic or mechanical means (including photocopying, recording,
or information storage and retrieval) without permission in writing from
the publisher.

Publisher’s Cataloging-in-Publication Data

Bertsekas, Dimitri P.
Network Optimization: Continuous and Discrete Models
Includes bibliographical references and index
1. Network analysis (Planning). 2. Mathematical Optimization. I. Title.
T57.85.B44 1998 658.4’032-dc20 98-70298

ISBN 1-886529-02-7
 ABOUT THE AUTHOR

Dimitri Bertsekas studied Mechanical and Electrical Engineering at

the National Technical University of Athens, Greece, and obtained his
Ph.D. in system science from the Massachusetts Institute of Technology.
He has held faculty positions at Stanford University and the Uni-
versity of Illinois. Since 1979 he has been teaching at the Massachusetts
Institute of Technology (M.I.T.), where he is currently McAfee Professor
of Engineering. He consults regularly with private industry and has held
editorial positions in several journals. His research spans several fields,
including optimization, control, large-scale computation, and data commu-
nication networks. He has written many research papers and he is the
author or coauthor of thirteen textbooks and research monographs.
Professor Bertsekas was awarded the INFORMS 1997 Prize for Re-
search Excellence in the Interface Between Operations Research and Com-
puter Science for his book ”Neuro-Dynamic Programming” (co-authored
with John Tsitsiklis), the 2000 Greek National Award for Operations Re-
search, and the 2001 ACC John R. Ragazzini Education Award. In 2001,
he was elected to the United States National Academy of Engineering.

iii
 ATHENA SCIENTIFIC
OPTIMIZATION AND COMPUTATION SERIES

1. Convex Analysis and Optimization, by Dimitri P. Bertsekas, with

Angelia Nedić and Asuman E. Ozdaglar, 2003, ISBN 1-886529-
45-0, 560 pages
2. Introduction to Probability by Dimitri P. Bertsekas and John
Tsitsiklis, 2002, ISBN 1-886529-40-X, 430 pages
3. Dynamic Programming and Optimal Control, Vols. I and II, 2nd
Edition, by Dimitri P. Bertsekas, 2001, ISBN 1-886529-08-6, 704
pages
4. Nonlinear Programming, 2nd Edition, by Dimitri P. Bertsekas,
1999, ISBN 1-886529-00-0, 800 pages
5. Network Optimization: Continuous and Discrete Models by Dim-
itri P. Bertsekas, 1998, ISBN 1-886529-02-7, 608 pages
6. Network Flows and Monotropic Optimization by R. Tyrrell Rock-
afellar, 1998, ISBN 1-886529-06-X, 634 pages
7. Introduction to Linear Optimization by Dimitris Bertsimas and
John N. Tsitsiklis, 1997, ISBN 1-886529-19-1, 608 pages
8. Parallel and Distributed Computation: Numerical Methods by
Dimitri P. Bertsekas and John N. Tsitsiklis, 1997, ISBN 1-886529-
01-9, 718 pages
9. Neuro-Dynamic Programming, by Dimitri P. Bertsekas and John
N. Tsitsiklis, 1996, ISBN 1-886529-10-8, 512 pages
10. Constrained Optimization and Lagrange Multiplier Methods, by
Dimitri P. Bertsekas, 1996, ISBN 1-886529-04-3, 410 pages
11. Stochastic Optimal Control: The Discrete-Time Case by Dimitri
P. Bertsekas and Steven E. Shreve, 1996, ISBN 1-886529-03-5,
330 pages

iv
Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . p. 1
1.1. Graphs and Flows . . . . . . . . . . . . . . . . . . . . p. 3
1.1.1. Paths and Cycles . . . . . . . . . . . . . . . . . . p. 4
1.1.2. Flow and Divergence . . . . . . . . . . . . . . . . p. 6
1.1.3. Path Flows and Conformal Decomposition . . . . . . . p. 7
1.2. Network Flow Models – Examples . . . . . . . . . . . . . p. 8
1.2.1. The Minimum Cost Flow Problem . . . . . . . . . . p. 9
1.2.2. Network Flow Problems with Convex Cost . . . . . . . p. 16
1.2.3. Multicommodity Flow Problems . . . . . . . . . . . p. 17
1.2.4. Discrete Network Optimization Problems . . . . . . . p. 19
1.3. Network Flow Algorithms – An Overview . . . . . . . . . p. 20
1.3.1. Primal Cost Improvement . . . . . . . . . . . . . . p. 21
1.3.2. Dual Cost Improvement . . . . . . . . . . . . . . . p. 24
1.3.3. Auction . . . . . . . . . . . . . . . . . . . . . . p. 27
1.3.4. Good, Bad, and Polynomial Algorithms . . . . . . . . p. 35
1.4. Notes, Sources, and Exercises . . . . . . . . . . . . . . . p. 37

2. Shortest Path Problems . . . . . . . . . . . . . . . p. 51

2.1. Problem Formulation and Applications . . . . . . . . . . p. 52
2.2. A Generic Shortest Path Algorithm . . . . . . . . . . . . p. 57
2.3. Label Setting (Dijkstra) Methods . . . . . . . . . . . . . p. 65
2.3.1. Performance of Label Setting Methods . . . . . . . . . p. 68
2.3.2. The Binary Heap Method . . . . . . . . . . . . . . p. 69
2.3.3. Dial’s Algorithm . . . . . . . . . . . . . . . . . . p. 70
2.4. Label Correcting Methods . . . . . . . . . . . . . . . . p. 73
2.4.1. The Bellman-Ford Method . . . . . . . . . . . . . . p. 73
2.4.2. The D’Esopo-Pape Algorithm . . . . . . . . . . . . p. 75
2.4.3. The SLF and LLL Algorithms . . . . . . . . . . . . p. 76
2.4.4. The Threshold Algorithm . . . . . . . . . . . . . . p. 78
2.4.5. Comparison of Label Setting and Label Correcting . . . p. 80
2.5. Single Origin/Single Destination Methods . . . . . . . . . p. 81
2.5.1. Label Setting . . . . . . . . . . . . . . . . . . . . p. 81

v
 vi Contents

2.5.2. Label Correcting . . . . . . . . . . . . . . . . . . p. 81

2.6. Auction Algorithms . . . . . . . . . . . . . . . . . . . p. 86
2.7. Multiple Origin/Multiple Destination Methods . . . . . . . p. 96
2.8. Notes, Sources, and Exercises . . . . . . . . . . . . . . . p. 98

3. The Max-Flow Problem . . . . . . . . . . . . . . p. 115

3.1. The Max-Flow and Min-Cut Problems . . . . . . . . . . p. 116
3.1.1. Cuts in a Graph . . . . . . . . . . . . . . . . . p. 119
3.1.2. The Max-Flow/Min-Cut Theorem . . . . . . . . . . p. 121
3.1.3. The Maximal and Minimal Saturated Cuts . . . . . . p. 123
3.1.4. Decomposition of Infeasible Network Problems . . . . p. 124
3.2. The Ford-Fulkerson Algorithm . . . . . . . . . . . . . p. 125
3.3. Price-Based Augmenting Path Algorithms . . . . . . . . p. 132
3.3.1. A Price-Based Path Construction Algorithm . . . . . p. 135
3.3.2. A Price-Based Max-Flow Algorithm . . . . . . . . . p. 139
3.4. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 139

4. The Min-Cost Flow Problem . . . . . . . . . . . . p. 151

4.1. Transformations and Equivalences . . . . . . . . . . . . p. 152
4.1.1. Setting the Lower Flow Bounds to Zero . . . . . . . p. 152
4.1.2. Eliminating the Upper Flow Bounds . . . . . . . . . p. 153
4.1.3. Reduction to a Circulation Format . . . . . . . . . p. 154
4.1.4. Reduction to an Assignment Problem . . . . . . . . p. 154
4.2. Duality . . . . . . . . . . . . . . . . . . . . . . . p. 155
4.2.1. Interpretation of CS and the Dual Problem . . . . . . p. 162
4.2.2. Duality and CS for Nonnegativity Constraints . . . . p. 163
4.3. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 164

5. Simplex Methods for Min-Cost Flow . . . . . . . . . p. 169

5.1. Main Ideas in Simplex Methods . . . . . . . . . . . . . p. 170
5.1.1. Using Prices to Obtain the In-Arc . . . . . . . . . . p. 176
5.1.2. Obtaining the Out-Arc . . . . . . . . . . . . . . p. 179
5.1.3. Dealing with Degeneracy . . . . . . . . . . . . . . p. 183
5.2. The Basic Simplex Algorithm . . . . . . . . . . . . . . p. 186
5.2.1. Termination Properties of the Simplex Method . . . . p. 187
5.2.2. Initialization of the Simplex Method . . . . . . . . . p. 188
5.3. Extension to Problems with Upper and Lower Bounds . . . p. 195
5.4. Implementation Issues . . . . . . . . . . . . . . . . . p. 199
5.5. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 203

6. Dual Ascent Methods for Min-Cost Flow . . . . . . . p. 213

6.1. Dual Ascent . . . . . . . . . . . . . . . . . . . . . p. 214
 Contents vii

6.2. The Primal-Dual (Sequential Shortest Path) Method . . . p. 221

6.3. The Relaxation Method . . . . . . . . . . . . . . . . p. 234
6.4. Solving Variants of an Already Solved Problem . . . . . . p. 243
6.5. Implementation Issues . . . . . . . . . . . . . . . . . p. 243
6.6. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 244

7. Auction Algorithms for Min-Cost Flow . . . . . . . p. 251

7.1. The Auction Algorithm for the Assignment Problem . . . p. 252
7.1.1. The Main Auction Algorithm . . . . . . . . . . . . p. 253
7.1.2. Approximate Coordinate Descent Interpretation . . . p. 257
7.1.3. Variants of the Auction Algorithm . . . . . . . . . p. 257
7.1.4. Computational Complexity –
-Scaling . . . . . . . . p. 259
7.1.5. Dealing with Infeasibility . . . . . . . . . . . . . p. 265
7.2. Extensions of the Auction Algorithm . . . . . . . . . . p. 268
7.2.1. Reverse Auction . . . . . . . . . . . . . . . . . p. 268
7.2.2. Auction Algorithms for Asymmetric Assignment . . . p. 272
7.2.3. Auction Algorithms with Similar Persons . . . . . . p. 279
7.3. The Preflow-Push Algorithm for Max-Flow . . . . . . . . p. 282
7.3.1. Analysis and Complexity . . . . . . . . . . . . . . p. 285
7.3.2. Implementation Issues . . . . . . . . . . . . . . . p. 293
7.3.3. Relation to the Auction Algorithm . . . . . . . . . p. 294
7.4. The
-Relaxation Method . . . . . . . . . . . . . . . p. 304
7.4.1. Computational Complexity –
-Scaling . . . . . . . . p. 310
7.4.2. Implementation Issues . . . . . . . . . . . . . . . p. 318
7.5. The Auction/Sequential Shortest Path Algorithm . . . . . p. 320
7.6. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 326

8. Nonlinear Network Optimization . . . . . . . . . . p. 337

8.1. Convex and Separable Problems . . . . . . . . . . . . p. 339
8.2. Problems with Side Constraints . . . . . . . . . . . . . p. 346
8.3. Multicommodity Flow Problems . . . . . . . . . . . . p. 349
8.4. Integer Constraints . . . . . . . . . . . . . . . . . . p. 355
8.5. Networks with Gains . . . . . . . . . . . . . . . . . p. 360
8.6. Optimality Conditions . . . . . . . . . . . . . . . . . p. 365
8.7. Duality . . . . . . . . . . . . . . . . . . . . . . . p. 370
8.8. Algorithms and Approximations . . . . . . . . . . . . p. 375
8.8.1. Feasible Direction Methods . . . . . . . . . . . . . p. 375
8.8.2. Piecewise Linear Approximation . . . . . . . . . . p. 380
8.8.3. Interior Point Methods . . . . . . . . . . . . . . p. 382
8.8.4. Penalty and Augmented Lagrangian Methods . . . . . p. 384
8.8.5. Proximal Minimization . . . . . . . . . . . . . . p. 386
8.8.6. Smoothing . . . . . . . . . . . . . . . . . . . . p. 387
8.8.7. Transformations . . . . . . . . . . . . . . . . . p. 389
 viii Contents

8.9. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 398

9. Convex Separable Network Problems . . . . . . . . p. 407

9.1. Convex Functions of a Single Variable . . . . . . . . . . p. 408
9.2. Optimality Conditions . . . . . . . . . . . . . . . . . p. 412
9.3. Duality . . . . . . . . . . . . . . . . . . . . . . . p. 414
9.4. Dual Function Differentiability . . . . . . . . . . . . . p. 426
9.5. Algorithms for Differentiable Dual Problems . . . . . . . p. 430
9.6. Auction Algorithms . . . . . . . . . . . . . . . . . . p. 433
9.6.1. The
-Relaxation Method . . . . . . . . . . . . . p. 441
9.6.2. Auction/Sequential Shortest Path Algorithm . . . . . p. 446
9.7. Monotropic Programming . . . . . . . . . . . . . . . p. 449
9.8. Notes, Sources, and Exercises . . . . . . . . . . . . . . p. 463

10. Network Problems with Integer Constraints . . . . . p. 467

10.1. Formulation of Integer-Constrained Problems . . . . . . p. 469
10.2. Branch-and-Bound . . . . . . . . . . . . . . . . . . p. 483
10.3. Lagrangian Relaxation . . . . . . . . . . . . . . . . p. 492
10.3.1. Subgradients of the Dual Function . . . . . . . . . p. 497
10.3.2. Subgradient Methods . . . . . . . . . . . . . . . p. 499
10.3.3. Cutting Plane Methods . . . . . . . . . . . . . . p. 503
10.3.4. Decomposition and Multicommodity Flows . . . . . p. 507
10.4. Local Search Methods . . . . . . . . . . . . . . . . p. 512
10.4.1. Genetic Algorithms . . . . . . . . . . . . . . . p. 514
10.4.2. Tabu Search . . . . . . . . . . . . . . . . . . . p. 515
10.4.3. Simulated Annealing . . . . . . . . . . . . . . . p. 516
10.5. Rollout Algorithms . . . . . . . . . . . . . . . . . . p. 517
10.6. Notes, Sources, and Exercises . . . . . . . . . . . . . p. 525

Appendix A: Mathematical Review . . . . . . . . . . p. 545

A.1. Sets . . . . . . . . . . . . . . . . . . . . . . . . p. 546
A.2. Euclidean Space . . . . . . . . . . . . . . . . . . . p. 547
A.3. Matrices . . . . . . . . . . . . . . . . . . . . . . p. 547
A.4. Analysis . . . . . . . . . . . . . . . . . . . . . . . p. 548
A.5. Convex Sets and Functions . . . . . . . . . . . . . . p. 551
A.6. Subgradients . . . . . . . . . . . . . . . . . . . . . p. 553

References . . . . . . . . . . . . . . . . . . . . . . . . . p. 555

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 587
Preface

Network optimization lies in the middle of the great divide that separates
the two major types of optimization problems, continuous and discrete.
The ties between linear programming and combinatorial optimization can
be traced to the representation of the constraint polyhedron as the convex
hull of its extreme points. When a network is involved, however, these ties
become much stronger because the extreme points of the polyhedron are in-
teger and represent solutions of combinatorial problems that are seemingly
unrelated to linear programming. Because of this structure and also be-
cause of their intuitive character, network models provide ideal vehicles for
explaining many of the fundamental ideas in both continuous and discrete
optimization.
Aside from their interesting methodological characteristics, network
models are also used extensively in practice, in an ever expanding spec-
trum of applications. Indeed collectively, network problems such as short-
est path, assignment, max-flow, transportation, transhipment, spanning
tree, matching, traveling salesman, generalized assignment, vehicle rout-
ing, and multicommodity flow constitute the most common class of practi-
cal optimization problems. There has been steady progress in the solution
methodology of network problems, and in fact the progress has accelerated
in the last fifteen years thanks to algorithmic and technological advances.
The purpose of this book is to provide a fairly comprehensive and up-
to-date development of linear, nonlinear, and discrete network optimization
problems. The interplay between continuous and discrete structures has
been highlighted, the associated analytical and algorithmic issues have been
treated quite extensively, and a guide to important network models and
applications has been provided.
Regarding continuous network optimization, we focus on two ideas,
which are also fundamental in general mathematical programming: dual-
ity and iterative cost improvement. We provide an extensive treatment of
iterative algorithms for the most common linear cost problem, the mini-
mum cost flow or transhipment problem, and for its convex cost extensions.
The discussion of duality is comprehensive: it starts with linear network

ix
x Preface

programming duality, and culminates with Rockafellar’s development of

monotropic programming duality.
Regarding discrete network optimization, we illustrate problem for-
mulation through major paradigms such as traveling salesman, generalized
assignment, spanning tree, matching, and routing. This is essential because
the structure of discrete optimization problems is far less streamlined than
the structure of their continuous counterparts, and familiarity with impor-
tant types of problems is important for modeling, analysis, and algorith-
mic solution. We also develop the main algorithmic approaches, including
branch-and-bound, Lagrangian relaxation, Dantzig-Wolfe decomposition,
heuristics, and local search methods.
This is meant to be an introductory book that covers a very broad
variety of topics. It is thus inevitable that some topics have been treated in
less detail than others. The choices made reflect in part personal taste and
expertise, and in part a preference for simple models that can help most
effectively the reader develop insight. At the same time, our analysis and
presentation aims to enhance the reader’s mathematical modeling ability in
two ways: by delineating the range of problems for which various algorithms
are applicable and efficient, and by providing many examples of problem
formulation.
The chapter-by-chapter description of the book follows:
Chapter 1: This is an introductory chapter that establishes terminology
and basic notions about graphs, discusses some examples of network mod-
els, and provides some orientation regarding linear network optimization
algorithms.
Chapter 2: This chapter provides an extensive treatment of shortest path
problems. It covers the major methods, and discusses their theoretical and
practical performance.
Chapter 3: This chapter focuses on the max-flow problem and develops
the class of augmenting path algorithms for its solution. In addition to the
classical variants of the Ford-Fulkerson method, a recent algorithm based
on auction ideas is discussed.
Chapter 4: The minimum cost flow problem (linear cost, single commod-
ity, no side constraints) and its equivalent variants are introduced here.
Subsequently, the basic duality theory for the problem is developed and
interpreted.
Chapter 5: This chapter focuses on simplex methods for the minimum
cost flow problem. The basic results regarding the integrality of solutions
are developed here constructively, using the simplex method. Furthermore,
the duality theory of Chapter 4 is significantly strengthened.
Chapter 6: This chapter develops dual ascent methods, including primal-
dual, sequential shortest path, and relaxation methods.
Preface xi

Chapter 7: This chapter starts with the auction algorithm for the assign-
ment problem, and proceeds to show how this algorithm can be extended
to more complex problems. In this way, preflow-push methods for the
max-flow problem and the
-relaxation method for the minimum cost flow
problem are obtained. Several additional variants of auction algorithms
are developed.
Chapter 8: This is an important chapter that marks the transition from
linear to nonlinear network optimization. The primary focus is on continu-
ous (convex) problems, and their associated broad variety of structures and
methodology. In particular, there is an overview of the types of algorithms
from nonlinear programming that are useful in connection with various con-
vex network problems. There is also some discussion of discrete (integer)
problems with an emphasis on their ties with continuous problems.
Chapter 9: This is a fairly sophisticated chapter that is directed primar-
ily towards the advanced and/or research-oriented reader. It deals with
separable convex problems, discusses their connection with classical net-
work equilibrium problems, and develops their rich theoretical structure.
The salient features of this structure are a particularly sharp duality the-
ory, and a combinatorial connection of descent directions with the finite
set of elementary vectors of the subspace defined by the conservation of
flow constraints. Besides treating convex separable network problems, this
chapter provides an introduction to monotropic programming, which is the
largest class of nonlinear programming problems that possess the strong
duality and combinatorial properties of linear programs. This chapter also
develops auction algorithms for convex separable problems and provides an
analysis of their running time.
Chapter 10: This chapter deals with the basic methodological approaches
for integer-constrained problems. There is a treatment of exact methods
such as branch-and-bound, and the associated methods of Lagrangian re-
laxation, subgradient optimization, and cutting plane. There is also a
description of approximate methods based on local search, such as genetic
algorithms, tabu search, and simulated annealing. Finally, there is a dis-
cussion of rollout algorithms, a relatively new and broadly applicable class
of approximate methods, which can be used in place of, or in conjunction
with local search.
The book can be used for a course on network optimization or for part
of a course on introductory optimization at the first-year graduate level.
With the exception of some of the material in Chapter 9, the prerequisites
are fairly elementary. The main one is a certain degree of mathematical
maturity, as provided for example by a rigorous mathematics course beyond
the calculus level. One may cover most of the book in a course on linear
and nonlinear network optimization. A shorter version of this course may
consist of Chapters 1-5, and 8. Alternatively, one may teach a course that
xii Preface

focuses on linear and discrete network optimization, using Chapters 1-5,

a small part of Chapter 8, and Chapter 10. Actually, in these chapter
sequences, it is not essential to cover Chapter 5, if one is content with
weaker versions of duality results (given in Chapter 4) and one establishes
the integrality properties of optimal solutions with a line of argument such
as the one given in Exercise 1.34. The following figure illustrates the chapter
dependencies.

Chapters 1-5
(Intro/Linear)

Chapter 6 Chapter 7 Chapter 8 Chapter 10

(Dual Methods) (Auction) (Nonlinear/Discrete) (Integer)

Chapter 9
(Convex)

The book contains a large number of examples and exercises, which

should enhance its suitability for classroom instruction. Some of the exer-
cises are theoretical in nature and supplement substantially the main text.
Solutions to a subset of these (as well as errata and additional material)
will be posted and periodically updated on the book’s web page:
http://www.athenasc.com/netsbook.html
Also, the author’s web page
http://web.mit.edu/dimitrib/www/home.html
contains listings of FORTRAN codes implementing many of the algorithms
discussed in the book.
There is a very extensive literature on continuous and discrete net-
work optimization, and to give a complete bibliography and a historical
account of the research that led to the present form of the subject would
have been impossible. Thus I have not attempted to compile a compre-
hensive list of original contributions to the field. I have cited sources that
I have used extensively, that provide important extensions to the material
of the book, that survey important topics, or that are particularly well
suited for further reading. I have also cited selectively a few sources that
are historically significant, but the reference list is far from exhaustive in
this respect. Generally, to aid researchers in the field, I have preferred to
cite surveys and textbooks for subjects that are relatively mature, and to
Preface xiii

give a larger number of references for relatively recent developments.

A substantial portion of this book is based on the author’s research
on network optimization over the last twenty years. I was fortunate to
have several outstanding collaborators in this research, and I would like
to mention those with whom I have worked extensively. Eli Gafni assisted
with the computational experimentation using the auction algorithm and
the relaxation method for assignment problems in 1979. The idea of
-
scaling arose during my interactions with Eli at that time. Furthermore,
Eli collaborated extensively with me on various routing methods for data
networks, including projection methods for convex multicommodity flow
problems. Paul Tseng worked with me on network optimization starting
in 1982. Together we developed the RELAX codes, we developed several
extensions to the basic relaxation method and we collaborated closely on a
broad variety of other subjects, including the recent auction algorithms for
convex network problems and network problems with gains. David Cas-
tanon has worked extensively with me on a broad variety of algorithms
for assignment, transportation, and minimum cost flow problems, for both
serial and parallel computers, since 1987. John Tsitsiklis has been my coau-
thor and close collaborator for many years on a variety of optimization and
large scale computation topics, including some that deal with networks.
In addition to Eli, Paul, David, and John, I have had substantial research
collaborations with several colleagues, the results of which have been re-
flected in this book. In this regard, I would like to mention Jon Eckstein,
Bob Gallager, Francesca Guerriero, Roberto Musmanno, Stefano Pallot-
tino, and Maria-Grazia Scutellà. Several colleagues proofread portions of
the book, and contributed greatly with their suggestions. David Castanon,
Stefano Pallottino, Steve Patek, Serap Savari, Paul Tseng, and John Tsit-
siklis were particularly helpful in this regard. The research support of NSF
under grants from the DDM and the CCI divisions are very much appreci-
ated. My family has been a source of stability and loving support, without
which the book would not have been written.

Dimitri P. Bertsekas
Cambridge, Mass.
Spring 1998
1

Introduction

Contents

1.1. Graphs and Flows

1.1.1. Paths and Cycles
1.1.2. Flow and Divergence
1.1.3. Path Flows and Conformal Decomposition
1.2. Network Flow Models – Examples
1.2.1. The Minimum Cost Flow Problem
1.2.2. Network Flow Problems with Convex Cost
1.2.3. Multicommodity Flow Problems
1.2.4. Discrete Network Optimization Problems
1.3. Network Flow Algorithms – An Overview
1.3.1. Primal Cost Improvement
1.3.2. Dual Cost Improvement
1.3.3. Auction
1.3.4. Good, Bad, and Polynomial Algorithms
1.4. Notes, Sources, and Exercises

1
2 Introduction Chap. 1

Network flow problems are one of the most important and most frequently
encountered class of optimization problems. They arise naturally in the
analysis and design of large systems, such as communication, transporta-
tion, and manufacturing networks. They can also be used to model impor-
tant classes of combinatorial problems, such as assignment, shortest path,
and traveling salesman problems.
Loosely speaking, network flow problems consist of supply and de-
mand points, together with several routes that connect these points and
are used to transfer the supply to the demand. These routes may contain
intermediate transhipment points. Often, the supply, demand, and tran-
shipment points can be modeled by the nodes of a graph, and the routes can
be modeled by the paths of the graph. Furthermore, there may be multiple
“types” of supply/demand (or “commodities”) sharing the routes. There
may also be some constraints on the characteristics of the routes, such as
their carrying capacities, and some costs associated with using particu-
lar routes. Such situations are naturally modeled as network optimization
problems whereby, roughly speaking, we try to select routes that minimize
the cost of transfer of the supply to the demand.
This book deals with a broad spectrum of network optimization prob-
lems, involving linear and nonlinear cost functions. We pay special atten-
tion to four major classes of problems:
(a) The transhipment or minimum cost flow problem, which involves a
single commodity and a linear cost function. This problem has several
important special cases, such as the shortest path, the max-flow, the
assignment, and the transportation problems.
(b) The single commodity network flow problem with convex cost. This
problem is identical to the preceding transhipment problem, except
that the cost function is convex rather than linear.
(c) The multicommodity network flow problem with linear or convex cost.
This problem generalizes the preceding two classes of problems to the
case of multiple commodities.
(d) Discrete network optimization problems. These are problems where
the quantities transferred along the routes of the network are re-
stricted to take one of a finite number of values. Many combinatorial
optimization problems can be modeled in this way, including some
problems where the network structure is not immediately apparent.
Some discrete optimization problems are computationally very diffi-
cult, and in practice can only be solved approximately. Their algorith-
mic solution often involves the solution of “continuous” subproblems
that belong to the preceding three classes.
All of the network flow problems above can be mathematically mod-
eled in terms of graph-related notions. In Section 1.1, we introduce the
associated notation and terminology. In Section 1.2, we provide mathe-
 Sec. 1.1 Graphs and Flows 3

matical formulations and practical examples of network optimization mod-

els. Finally, in Section 1.3, we give an overview of some of the types of
computational algorithms that we develop in subsequent chapters.

1.1 GRAPHS AND FLOWS

In this section, we introduce some of the basic definitions relating to graphs,

paths, flows, and other related notions. Graph concepts are fairly intuitive,
and can be understood in terms of suggestive figures, but often involve
hidden subtleties. Thus the reader may wish to revisit the present section
and pay close attention to some of the fine points of the definitions.
A directed graph, G = (N , A), consists of a set N of nodes and a set
A of pairs of distinct nodes from N called arcs. The numbers of nodes and
arcs are denoted by N and A, respectively, and it is assumed throughout
that 1 ≤ N < ∞ and 0 ≤ A < ∞. An arc (i, j) is viewed as an ordered
pair, and is to be distinguished from the pair (j, i). If (i, j) is an arc, we
say that (i, j) is outgoing from node i and incoming to node j; we also say
that j is an outward neighbor of i and that i is an inward neighbor of j. We
say that arc (i, j) is incident to i and to j, and that i is the start node and
j is the end node of the arc. We also say that i and j are the end nodes of
arc (i, j). The degree of a node i is the number of arcs that are incident to
i. A graph is said to be complete if it contains all possible arcs; that is, if
there exists an arc for each ordered pair of nodes.
We do not exclude the possibility that there is a separate arc connect-
ing a pair of nodes in each of the two directions. However, we do not allow
more than one arc between a pair of nodes in the same direction, so that we
can refer unambiguously to the arc with start i and end j as arc (i, j). This
is done for notational convenience.† Our analysis can be simply extended
to handle multiple arcs with start i and end j; the extension is based on
modifying the graph by introducing for each such arc, an additional node,
call it n, together with the two arcs (i, n) and (n, j). On occasion, we will
pause to provide examples of this type of extension.
We note that much of the literature of graph theory distinguishes
between directed graphs where an arc (i, j) is an ordered pair to be distin-
guished from arc (j, i), and undirected graphs where an arc is associated
with a pair of nodes regardless of order. One may use directed graphs, even
in contexts where the use of undirected graphs would be appropriate and
conceptually simpler. For this, one may need to replace an undirected arc
(i, j) with two directed arcs (i, j) and (j, i) having identical characteristics.

† Some authors use a single symbol, such as a, to denote an arc, and use
something like s(a) and e(a) to denote the start and end nodes of a, respectively.
This notational method allows the existence of multiple arcs with the same start
and end nodes, but is also more cumbersome and less suggestive.
4 Introduction Chap. 1

We have chosen to deal exclusively with directed graphs because in our

development there are only a few occasions where undirected graphs are
convenient. Thus, all our references to a graph implicitly assume that the
graph is directed . In fact we often omit the qualifier “directed” and refer
to a directed graph simply as a graph.

1.1.1 Paths and Cycles

A path P in a directed graph is a sequence of nodes (n1 , n2 , . . . , nk ) with

k ≥ 2 and a corresponding sequence of k−1 arcs such that the ith arc in the
sequence is either (ni , ni+1 ) (in which case it is called a forward arc of the
path) or (ni+1 , ni ) (in which case it is called a backward arc of the path).
Nodes n1 and nk are called the start node (or origin) and the end node (or
destination) of P , respectively. A path is said to be forward (or backward )
if all of its arcs are forward (respectively, backward) arcs. We denote by
P + and P − the sets of forward and backward arcs of P , respectively.
A cycle is a path for which the start and end nodes are the same. A
path is said to be simple if it contains no repeated arcs and no repeated
nodes, except that the start and end nodes could be the same (in which
case the path is called a simple cycle). A Hamiltonian cycle is a simple
forward cycle that contains all the nodes of the graph. These definitions
are illustrated in Fig. 1.1. We mention that some authors use a slightly
different terminology: they use the term “walk” to refer to a path and they
use the term “path” to refer to a simple path.
Note that the sequence of nodes (n1 , n2 , . . . , nk ) is not sufficient to
specify a path; the sequence of arcs may also be important, as Fig. 1.1(c)
shows. The difficulty arises when for two successive nodes ni and ni+1 of
the path, both (ni , ni+1 ) and (ni+1 , ni ) are arcs, so there is ambiguity as
to which of the two is the corresponding arc of the path. If a path is known
to be forward or is known to be backward, it is uniquely specified by the
sequence of its nodes. Otherwise, however, the intended sequence of arcs
must be explicitly defined.
A graph that contains no simple cycles is said to be acyclic. A graph
is said to be connected if for each pair of nodes i and j, there is a path
starting at i and ending at j; it is said to be strongly connected if for each
pair of nodes i and j, there is a forward path starting at i and ending
at j. Thus, for example, the graph of Fig. 1.1(b) is connected but not
strongly connected. It can be shown that if a graph is connected and each
of its nodes has even degree, there is a cycle (not necessarily forward) that
contains all the arcs of the graph exactly once (see Exercise 1.5). Such
a cycle is called an Euler cycle, honoring the historically important work
of Euler; see the discussion in Section 10.1 about the Königsberg bridge
problem. Figure 1.2 gives an example of an Euler cycle.
We say that a graph G
= (N
, A
) is a subgraph of a graph G = (N , A)
if N ⊂ N and A
⊂ A. A tree is a connected acyclic graph. A spanning


Sec. 1.1 Graphs and Flows 5

Start Node n1 n2 n3 n4 End Node

(a) A simple forward path P = (n 1, n 2 , n 3 , n 4 ).

n1 n2

Set of backward arcs C-

n3
Set of forward arcs C +

(b) A simple cycle C = (n 1, n 2 , n 3 , n 1) which is neither forward nor backward.

Start Node n1 n2 n3 n4 n5 End Node

(c) Path P = (n 1, n 2 , n 3 , n 4 , n 5) with corresponding sequence of arcs

{ (n 1, n 2) , ( n 3 , n 2), ( n 3 , n 4), ( n 5 , n 4)}.

Figure 1.1: Illustration of various types of paths and cycles. The cycle in (b)
is not a Hamiltonian cycle; it is simple and contains all the nodes of the graph,
but it is not forward. Note that for the path (c), in order to resolve ambiguities,
it is necessary to specify the sequence of arcs of the path (rather than just the
sequence of nodes) because both (n3 , n4 ) and (n4 , n3 ) are arcs.

6
1 2 3 5 2

4 5 1 8

6 7 8 7 4
3

(a) (b) (c)

Figure 1.2: Example of an Euler cycle. Consider a 3 × 3 chessboard, where the

middle square has been deleted. A knight starting at one of the squares of the
board can visit every other square exactly once and return to the starting square
as shown in the graph (b), or equivalently in (c). In the process, the knight will
make all the possible moves (in one direction only), or equivalently, it will cross
every arc of the graph in (b) exactly once. The knight’s tour is an Euler cycle for
the graph of (b).
6 Introduction Chap. 1

tree of a graph G is a subgraph of G, which is a tree and includes all the

nodes of G. It can be shown [Exercise 1.14(c)] that a subgraph is a spanning
tree if and only if it is connected and it contains N − 1 arcs.

1.1.2 Flow and Divergence

In many applications involving graphs, it is useful to introduce a variable

that measures the quantity flowing through each arc, like for example,
electric current in an electric circuit, or water flow in a hydraulic network.
We refer to such a variable as the flow of an arc. Mathematically, the flow
of an arc (i, j) is simply a scalar (real number), which we usually denote
by xij . It is convenient to allow negative as well as positive values for flow.
In applications, a negative arc flow indicates that whatever is represented
by the flow (material, electric current, etc.), moves in a direction opposite
to the direction of the arc. We can always change the sign of a negative
arc flow to positive as long as we change the arc direction, so in many
situations we can assume without loss of generality that all arc flows are
nonnegative. For the development of a general methodology, however, this
device is often cumbersome, which is why we prefer to simply accept the
possibility of negative arc flows.

Given a graph (N , A), a set of flows xij | (i, j) ∈ A is referred to
as a flow vector . The divergence vector y associated with a flow vector x
is the N -dimensional vector with coordinates
 
yi = xij − xji , ∀ i ∈ N. (1.1)
{j|(i,j)∈A} {j|(j,i)∈A}

Thus, yi is the total flow departing from node i less the total flow arriving
at i; it is referred to as the divergence of i.
We say that node i is a source (respectively, sink ) for the flow vector
x if yi > 0 (respectively, yi < 0). If yi = 0 for all i ∈ N , then x is called
a circulation. These definitions are illustrated in Fig. 1.3. Note that by
adding Eq. (1.1) over all i ∈ N , we obtain

yi = 0.
i∈N

Every divergence vector y must satisfy this equation.

The flow vectors x that we will consider will often be constrained to
lie between given lower and upper bounds of the form

bij ≤ xij ≤ cij , ∀ (i, j) ∈ A.

Given a flow vector x that satisfies these bounds, we say that a path P is
unblocked with respect to x if, roughly speaking, we can send some positive
flow along P without violating the bound constraints; that is, if flow can
Sec. 1.1 Graphs and Flows 7

y2 = -2 (Sink)

x =1 2 x = -2
12 24
y 1 = 1 (Source) 1 x =1 x32= 0 4 y4 = 0 (Neither a source
23 nor a sink)
x =0 x34= 2
13 3
y 3 = 1 (Source)

(a)

y2 = 0

x =1 2 x 24= -1
12

y1 = 0 1 x =1 x 32= -1 4 y4 = 0
23

x13= -1 x34 = 1
3
y3 = 0

(b) A circulation

Figure 1.3: Illustration of flows xij and the corresponding divergences yi . The
flow in (b) is a circulation because yi = 0 for all i.

be increased on the set P + of the forward arcs of P , and can be decreased

on the set P − of the backward arcs of P :
xij < cij , ∀ (i, j) ∈ P + , bij < xij , ∀ (i, j) ∈ P − .
For example, in Fig. 1.3(a), suppose that all arcs (i, j) have flow bounds
bij = −2 and cij = 2. Then the path consisting of the sequence of nodes
(1, 2, 4) is unblocked, while the reverse path (4, 2, 1) is not unblocked.

1.1.3 Path Flows and Conformal Decomposition

A simple path flow is a flow vector that corresponds to sending a positive

amount of flow along a simple path; more precisely, it is a flow vector x
with components of the form
⎧
⎨a if (i, j) ∈ P + ,
xij = −a if (i, j) ∈ P − , (1.2)
⎩
0 otherwise,
where a is a positive scalar, and P + and P − are the sets of forward and
backward arcs, respectively, of some simple path P . Note that the path P
may be a cycle, in which case x is also called a simple cycle flow .
 8 Introduction Chap. 1

It is often convenient to break down a flow vector into the sum of

simple path flows. This leads to the notion of a conformal realization,
which we proceed to discuss.
We say that a path P conforms to a flow vector x if xij > 0 for all
forward arcs (i, j) of P and xij < 0 for all backward arcs (i, j) of P , and
furthermore either P is a cycle or else the start and end nodes of P are a
source and a sink of x, respectively. Roughly, a path conforms to a flow
vector if it “carries flow in the forward direction,” i.e., in the direction
from the start node to the end node. In particular, for a forward cycle to
conform to a flow vector, all its arcs must have positive flow. For a forward
path which is not a cycle to conform to a flow vector, its arcs must have
positive flow, and in addition the start and end nodes must be a source
and a sink, respectively; for example, in Fig. 1.3(a), the path consisting of
the sequence of arcs (1,2), (2,3), (3,4) does not conform to the flow vector
shown, because node 4, the end node of the path, is not a sink.
We say that a simple path flow xs conforms to a flow vector x if the
path P corresponding to xs via Eq. (1.2) conforms to x. This is equivalent
to requiring that
0 < xij for all arcs (i, j) with 0 < xsij ,
xij < 0 for all arcs (i, j) with xsij < 0,
and that either P is a cycle or else the start and end nodes of P are a
source and a sink of x, respectively.
An important fact is that any flow vector can be decomposed into a
set of conforming simple path flows, as illustrated in Fig. 1.4. We state
this as a proposition. The proof is based on an algorithm that can be used
to construct the conforming components one by one (see Exercise 1.2).

Proposition 1.1: (Conformal Realization Theorem) A nonzero

flow vector x can be decomposed into the sum of t simple path flow
vectors x1 , x2 , . . . , xt that conform to x, with t being at most equal to
the sum of the numbers of arcs and nodes A + N . If x is integer, then
x1 , x2 , . . . , xt can also be chosen to be integer. If x is a circulation,
then x1 , x2 , . . . , xt can be chosen to be simple cycle flows, and t ≤ A.

1.2 NETWORK FLOW MODELS – EXAMPLES

In this section we introduce some of the major classes of problems that will
be discussed in this book. We begin with the minimum cost flow problem,
which, together with its special cases, will be the subject of the following
six chapters.
Sec. 1.2 Network Flow Models – Examples 9

y2 = -2 (Sink)

x =1 2 x = -2
12 24

y 1 = 1 (Source) 1 x =1 x32= 0 4 y4 = 0 (Neither a source

23 nor a sink)
x =0 x34= 2
13 3
y 3 = 1 (Source)

Flow = 1 2 2 Flow = -1 2 Flow = -1

1 4 Flow = 1 4

3 Flow = 1 3 Flow = 1

Figure 1.4: Decomposition of a flow vector x into three simple path flows con-
forming to x. Consistent with the definition of conformance of a path flow, each
arc (i, j) of the three component paths carries positive (or negative) flow only if
xij > 0 (or xij < 0, respectively). The first two paths [(1, 2) and (3, 4, 2)] are not
cycles, but they start at a source and end at a sink, as required. Arcs (1, 3) and
(3, 2) do not belong to any of these paths because they carry zero flow. In this
example, the decomposition is unique, but in general this need not be the case.

1.2.1 The Minimum Cost Flow Problem

This problem is to find a set of arc flows that minimize a linear cost function,
subject to the constraints that they produce a given divergence vector and
they lie within some given bounds; that is,

minimize aij xij (1.3)
(i,j)∈A

subject to the constraints

 
xij − xji = si , ∀ i ∈ N, (1.4)
{j|(i,j)∈A} {j|(j,i)∈A}

bij ≤ xij ≤ cij , ∀ (i, j) ∈ A, (1.5)

where aij , bij , cij , and si are given scalars. We use the following terminol-
ogy:
aij : the cost coefficient (or simply cost) of (i, j),
bij and cij : the flow bounds of (i, j),
[bij , cij ]: the feasible flow range of (i, j),
10 Introduction Chap. 1

si : the supply of node i (when si is negative, the scalar −si is called

the demand of i).
We also refer to the constraints (1.4) and (1.5) as the conservation of flow
constraints, and the capacity constraints, respectively. A flow vector satis-
fying both of these constraints is called feasible, and if it satisfies just the
capacity constraints, it is called capacity-feasible. If there exists at least
one feasible flow vector, the minimum cost flow problem is called feasible;
otherwise it is called infeasible. On occasion, we will consider the variation
of the minimum cost flow problem where the lower or the upper flow bound
of some of the arcs is either −∞ or ∞, respectively. In these cases, we will
explicitly state so.
For a typical application of the minimum cost flow problem, think
of the nodes as locations (cities, warehouses, or factories) where a certain
product is produced or consumed. Think of the arcs as transportation
links between the locations, each with transportation cost aij per unit
transported. The problem then is to move the product from the production
points to the consumption points at minimum cost while observing the
capacity constraints of the transportation links.
However, the minimum cost flow problem has many applications that
are well beyond the transportation context just described, as will be seen
from the following examples. These examples illustrate how some impor-
tant discrete/combinatorial problems can be modeled as minimum cost flow
problems, and highlight the important connection between continuous and
discrete network optimization.

Example 1.1. The Shortest Path Problem

Suppose that each arc (i, j) of a graph is assigned a scalar cost aij , and suppose
that we define the cost of a forward path to be the sum of the costs of its
arcs. Given a pair of nodes, the shortest path problem is to find a forward
path that connects these nodes and has minimum cost. An analogy here is
made between arcs and their costs, and roads in a transportation network and
their lengths, respectively. Within this transportation context, the problem
becomes one of finding the shortest route between two geographical points.
Based on this analogy, the problem is referred to as the shortest path problem,
and the arc costs and path costs are commonly referred to as the arc lengths
and path lengths, respectively.
The shortest path problem arises in a surprisingly large number of con-
texts. For example in a data communication network, aij may denote the
average delay of a packet to cross the communication link (i, j), in which case
a shortest path is a minimum average delay path that can be used for routing
the packet from its origin to its destination. As another example, if pij is
the probability that a given arc (i, j) in a communication network is usable,
and each arc is usable independently of all other arcs, then the product of the
probabilities of the arcs of a path provides a measure of reliability of the path.
With this in mind, it is seen that finding the most reliable path connecting
Sec. 1.2 Network Flow Models – Examples 11

two nodes is equivalent to finding the shortest path between the two nodes
with arc lengths (− ln pij ).
The shortest path problem also arises often as a subroutine in algo-
rithms that solve other more complicated problems. Examples are the primal-
dual algorithm for solving the minimum cost flow problem (see Chapter 6),
and the conditional gradient and projection algorithms for solving multicom-
modity flow problems (see Chapter 8).
It is possible to cast the problem of finding a shortest path from node
s to node t as the following minimum cost flow problem:

minimize aij xij
(i,j)∈A

  1 if i = s,
(1.6)
subject to xij − xji = −1 if i = t,
{j|(i,j)∈A} {j|(j,i)∈A} 0 otherwise,
0 ≤ xij , ∀ (i, j) ∈ A.

To see this, let us associate with any forward path P from s to t the flow
vector x with components given by

1 if (i, j) belongs to P ,
xij = (1.7)
0 otherwise.

Then x is feasible for problem (1.6) and the cost of x is equal to the length
of P . Thus, if a vector x of the form (1.7) is an optimal solution of problem
(1.6), the corresponding path P is shortest.
Conversely, it can be shown that if problem (1.6) has at least one op-
timal solution, then it has an optimal solution of the form (1.7), with a
corresponding path P that is shortest. This is not immediately apparent, but
its proof can be traced to a remarkable fact that we will show in Chapter 5
about minimum cost flow problems with node supplies and arc flow bounds
that are integer: such problems, if they have an optimal solution, they have
an integer optimal solution, that is, a set of optimal arc flows that are integer
(an alternative proof of this fact is sketched in Exercise 1.34). From this it
follows that if problem (1.6) has an optimal solution, it has one with arc flows
that are 0 or 1, and which is of the form (1.7) for some path P . This path is
shortest because its length is equal to the optimal cost of problem (1.6), so it
must be less or equal to the cost of any other flow vector of the form (1.7),
and therefore also less or equal to the length of any other path from s to t.
Thus the shortest path problem is essentially equivalent with the minimum
cost flow problem (1.6).

Example 1.2. The Assignment Problem

Suppose that there are n persons and n objects that we have to match on a
one-to-one basis. There is a benefit or value aij for matching person i with
object j, and we want to assign persons to objects so as to maximize the total
12 Introduction Chap. 1

PERSONS OBJECTS
1 1
1 1

...

...
Figure 1.5: The graph represen-
1 a ij 1 tation of an assignment problem.
i j
...

...
1 1
n n

benefit. There is also a restriction that person i can be assigned to object j

only if (i, j) belongs to a given set of pairs A. Mathematically, we want to find
n A such that the objects
a set of person-object pairs (1, j1 ), . . . , (n, jn ) from
j1 , . . . , jn are all distinct, and the total benefit a is maximized.
i=1 iji
The assignment problem is important in many practical contexts. The
most obvious ones are resource allocation problems, such as assigning em-
ployees to jobs, machines to tasks, etc. There are also situations where the
assignment problem appears as a subproblem in methods for solving various
complex combinatorial problems (see Chapter 10).
We may associate any assignment with the set of variables {xij | (i, j) ∈
A}, where xij = 1 if person i is assigned to object j and xij = 0 otherwise.
The value of this assignment is (i,j)∈A aij xij . The restriction of one object

per person can be stated as x = 1 for all i and i xij = 1 for all j. We
j ij
may then formulate the assignment problem as the linear program

maximize aij xij
(i,j)∈A

subject to xij = 1, ∀ i = 1, . . . , n,
{j|(i,j)∈A} (1.8)

xij = 1, ∀ j = 1, . . . , n,
{i|(i,j)∈A}

0 ≤ xij ≤ 1, ∀ (i, j) ∈ A.

Actually we should further restrict xij to be either 0 or 1. However, as we

will show in Chapter 5, the above linear program has the property that if it
has a feasible solution at all, then it has an optimal solution where all xij
are either 0 or 1 (compare also with the discussion in the preceding example
and Exercise 1.34). In fact, the set of its optimal solutions includes all the
optimal assignments.
We now argue that the assignment/linear program (1.8) is a minimum
cost flow problem involving the graph shown in Fig. 1.5. Here, there are
2n nodes divided into two groups: n corresponding to persons and n corre-
sponding to objects. Also, for every possible pair (i, j) ∈ A, there is an arc
connecting person i with object j. The variable xij is the flow of arc (i, j).
Sec. 1.2 Network Flow Models – Examples 13

The constraint 
xij = 1
{j|(i,j)∈A}

indicates that the divergence of person/node i should be equal to 1, while the

constraint 
xij = 1
{i|(i,j)∈A}

indicates that the divergence of object/node j should be equal to -1. Finally,

we may view (−aij ) as the cost coefficient of the arc (i, j) (by reversing the
sign of aij , we convert the problem from a maximization to a minimization
problem).

Example 1.3. The Max-Flow Problem

In the max-flow problem, we have a graph with two special nodes: the source,
denoted by s, and the sink , denoted by t. Roughly, the objective is to move as
much flow as possible from s into t while observing the capacity constraints.
More precisely, we want to find a flow vector that makes the divergence of all
nodes other than s and t equal to 0 while maximizing the divergence of s.

All cost coefficients are

zero except for ats

Source s t Sink

Artificial feedback arc

Cost coefficient = -1

Figure 1.6: The minimum cost flow representation of a max-flow problem.

At the optimum, the flow xts equals the maximum flow that can be sent from
s to t through the subgraph obtained by deleting the artificial arc (t, s).

The max-flow problem arises in many practical contexts, such as calcu-

lating the throughput of a highway system or a communication network. It
also arises often as a subproblem in more complicated problems or algorithms;
in particular, it bears a fundamental connection to the question of existence of
a feasible solution of a general minimum cost flow problem (see our discussion
14 Introduction Chap. 1

in Chapter 3). Finally, several discrete/combinatorial optimization problems

can be formulated as max-flow problems (see the Exercises in Chapter 3).
We formulate the problem as a special case of the minimum cost flow
problem by assigning cost 0 to all arcs and by introducing an artificial arc
(t, s) with cost −1, as shown in Fig. 1.6. Mathematically, the problem is:

maximize xts
subject to
 
xij − xji = 0, ∀ i ∈ N with i = s and i = t,
{j|(i,j)∈A} {j|(j,i)∈A}
 
xsj = xit = xts ,
{j|(s,j)∈A} {i|(i,t)∈A}

bij ≤ xij ≤ cij , ∀ (i, j) ∈ A with (i, j) = (t, s).

Viewing the problem as a maximization is consistent with its intuitive inter-

pretation. Alternatively, we could write the problem as a minimization of
−xts subject to the same constraints. Also, we could introduce upper and
lower bounds on xts ,
 
bit ≤ xts ≤ cit ,
{i|(i,t)∈A} {i|(i,t)∈A}

but these bounds are actually redundant since they are implied by the other
upper and lower arc flow bounds.

Example 1.4. The Transportation Problem

This problem is the same as the assignment problem except that the node
supplies need not be 1 or −1, and the numbers of sources and sinks need not
be equal. It has the form

minimize aij xij
(i,j)∈A

subject to xij = αi , ∀ i = 1, . . . , m,
{j|(i,j)∈A} (1.9)

xij = βj , ∀ j = 1, . . . , n,
{i|(i,j)∈A}

0 ≤ xij ≤ min{αi , βj }, ∀ (i, j) ∈ A.

Here αi and βj are positive scalars, which for feasibility must satisfy


m 
n

αi = βj ,
i=1 j=1
Sec. 1.2 Network Flow Models – Examples 15

(add the conservation of flow constraints). In an alternative formulation,

the upper bound constraint xij ≤ min{αi , βj } could be discarded, since it is
implied by the conservation of flow and the nonnegativity constraints.
As a practical example of a transportation problem that has a combi-
natorial flavor, suppose that we have m communication terminals, each to be
connected to one of n traffic concentrators. We introduce variables xij , which
take the value 1 if terminal i is connected to concentrator j. Assuming that
concentrator j can be connected to no more than bj terminals, we obtain the
constraints

m

xij ≤ bj , ∀ j = 1, . . . , n.
i=1

Also, since each terminal must be connected to exactly one concentrator, we

have the constraints

n

xij = 1, ∀ i = 1, . . . , m.
j=1

Assuming that there is a cost aij for connecting terminal i to concentrator j,

the problem is to find the connection of minimum cost, that is, to minimize

m 
n

aij xij
i=1 j=1

subject to the preceding constraints. This problem is not yet a transportation

problem of the form (1.9) for two reasons:
(a) The arc flows xij are constrained to be 0 or 1.
m
(b) The constraints i=1
xij ≤ bj are not equality constraints, as required
in problem (1.9).
It turns out, however, that we can ignore the 0-1 constraint on xij . As
discussed in connection with the shortest path and assignment problems,
even if we relax this constraint and replace it with the capacity constraint
0 ≤ xij ≤ 1, there is an optimal solution such that each xij is either 0 or
1. Furthermore, to convert the inequality constraints to equalities, we can
n
introduce a total of b − m “dummy” terminals that can be connected
j=1 j
at zero cost to all of the concentrators. In particular, we introduce a special
supply node 0 together with the constraint

n 
n

x0j = bj − m,
j=1 j=1
n
and we change the inequality constraints j=1
xij ≤ bj to

m

x0j + xij = bj .
i=1

The resulting problem has the transportation structure of problem (1.9), and
is equivalent to the original problem.
16 Introduction Chap. 1

1.2.2 Network Flow Problems with Convex Cost

A more general version of the minimum cost flow problem arises when the
cost function is convex rather than linear. An important special case is the
problem

minimize fij (xij )
(i,j)∈A
 
subject to xij − xji = si , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

xij ∈ Xij , ∀ (i, j) ∈ A,

where fij is a convex function of the flow xij of arc (i, j), si are given
scalars, and Xij are convex intervals of real numbers, such as for example

Xij = [bij , cij ],

where bij and cij are given scalars. We refer to this as the separable convex
cost network flow problem, because the cost function separates into the sum
of cost functions, one per arc. This problem will be discussed in detail in
Chapters 8 and 9.

Example 1.5. The Matrix Balancing Problem

Here the problem is to find an m × n matrix X that has given row sums and
column sums, and approximates a given m × n matrix M in some optimal
manner. We can formulate such a problem in terms of a graph consisting of
m sources and n sinks. In this graph, the set of arcs consists of the pairs
(i, j) for which the corresponding entry xij of the matrix X is allowed to be
nonzero. The given row sums ri and the given column sums cj are expressed
as the constraints

xij = ri , i = 1, . . . , m,
{j|(i,j)∈A}


xij = cj , j = 1, . . . , n.
{i|(i,j)∈A}

There may be also bounds for the entries xij of X. Thus, the structure of
this problem is similar to the structure of a transportation problem. The cost
function to be optimized has the form

fij (xij ),
(i,j)∈A
Sec. 1.2 Network Flow Models – Examples 17

and expresses the objective of making the entries of X close to the corre-
sponding entries of the given matrix M . A commonly used example is the
quadratic function

fij (xij ) = wij (xij − mij )2 ,
(i,j)∈A

where wij are given positive scalars.

Another interesting cost function is the logarithmic

 
xij
fij (xij ) = xij ln −1 ,
mij

where we assume that mij > 0 for all (i, j) ∈ A. Note that this function is not
defined for xij ≤ 0, so to obtain a problem that fits our framework, we must
use a constraint interval of the form Xij = (0, ∞) or Xij = (0, cij ], where cij
is a positive scalar.
An example of a practical problem that can be addressed using the
preceding optimization model is to predict the distribution matrix X of tele-
phone traffic between m origins and n destinations. Here we are given the
total supplies ri of the origins and the total demands cj of the destinations,
and we are also given some matrix M that defines a nominal traffic pattern
obtained from historical data.

There are other types of network flow problems with convex cost that
often arise in practice. We generically represent such problems in the form

minimize f (x)
subject to x ∈ F

where F is a convex subset of flow vectors in a graph and f is a convex

function over the set F . We will discuss in some detail various classes
of problems of this type in Chapter 8, and we will see that they arise in
several different ways; for example, the cost function may be nonseparable
because of coupling of the costs of several arc flows, and/or there may be
side constraints, whereby the flows of several arcs are jointly restricted by
the availability of resource. An important example is multicommodity flow
problems, which we discuss next.

1.2.3 Multicommodity Flow Problems

Multicommodity network flow problems involve several flow “types” or

commodities, which simultaneously use the network and are coupled through
either the arc flow bounds, or through the cost function. Important exam-
ples of such problems arise in communication, transportation, and man-
ufacturing networks. For example, in communication networks the com-
modities are the streams of different classes of traffic (telephone, data,
18 Introduction Chap. 1

video, etc.) that involve different origin-destination pairs. Thus there is

a separate commodity per class of traffic and origin-destination pair. The
following example introduces this context. In Chapter 8, we will discuss
similar and/or more general multicommodity network flow problems that
arise in other practical contexts.

Example 1.6. Routing in Data Networks

We are given a directed graph, which is viewed as a model of a data com-

munication network. We are also given a set of ordered node pairs (im , jm ),
m = 1, . . . , M , referred to as origin-destination (OD) pairs. The nodes im
and jm are referred to as the origin and the destination of the OD pair. For
each OD pair (im , jm ), we are given a scalar rm that represents its input
traffic. In the context of routing of data in a communication network, rm
(measured for example in bits/second) is the arrival rate of traffic entering
the network at node im and exiting at node jm . The routing objective is to
divide each rm among the many paths from the origin im to the destination
jm in a way that the resulting total arc flow pattern minimizes a suitable cost
function (see Fig. 1.7).

Origin of Destination of
OD pair (im , jm ) OD pair (im , jm )
im jm
rm rm

Figure 1.7: Illustration of how the input rm of the OD pair (im , jm ) is

divided into nonnegative path flows that start at im and end at jm . The
flows of the different OD pairs interact by sharing the arcs of the network.

If we denote by xij (m) the flow on arc (i, j) of OD pair (im , jm ), we

have the conservation of flow constraints

  rm if i = im ,
xij (m) − xji (m) = −rm if i = jm , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A} 0 otherwise,

for each m = 1, . . . , M . Furthermore, the flows xij (m) are required to be

nonnegative, and possibly to satisfy additional constraints, such as upper
bounds. The cost function often has the form

f (x) = fij (yij ),
(i,j)∈A
Sec. 1.2 Network Flow Models – Examples 19

where fij is a function of the total flow of arc (i, j)


M

yij = xij (m).

m=1

Such a cost function is often based on a queueing model of average delay (see
for example the data network textbook by Bertsekas and Gallager [1992]).

1.2.4 Discrete Network Optimization Problems

Many linear or convex network flow problems, in addition to the conser-

vation of flow constraints and arc flow bounds, involve some additional
constraints. In particular, there may be constraints that couple the flows
of different arcs, and there may also be integer constraints on the arc flows,
such as for example that each arc flow be either 0 or 1. Several famous
combinatorial optimization problems, such as the following one, are of this
type.

Example 1.7. The Traveling Salesman Problem

This problem refers to a salesman who wants to find a minimum mileage/cost

tour that visits each of N given cities exactly once and returns to the city
he started from. To convert this to a network flow problem, we associate a
node with each city i = 1, . . . , N , and we introduce an arc (i, j) with traversal
cost aij for each ordered pair of nodes i and j. A tour is synonymous to
a Hamiltonian cycle, which was earlier defined to be a simple forward cycle
that contains all the nodes of the graph. Equivalently, a tour is a connected
subgraph that consists of N arcs, such that there is exactly one incoming and
one outgoing arc for each node i = 1, . . . , N . The problem is to find a tour
with minimum sum of arc costs.
To formulate this problem as a network flow problem, we denote by xij
the flow of arc (i, j) and we require that this flow is either 1 or 0, indicating
that the arc is or is not part of the tour, respectively. The cost of a tour T is
then 
aij xij .
(i,j)∈T

The constraint that each node has a single incoming and a single outgoing arc
on the tour is expressed by the following two conservation of flow equations:

xij = 1, i = 1, . . . , N,
j=1,...,N
j=i


xij = 1, j = 1, . . . , N.
i=1,...,N
i=j
 20 Introduction Chap. 1

There is one additional connectivity constraint:

the subgraph with node set N and arc set {(i, j) | xij = 1} is connected.

If this constraint was not present, the problem would be an ordinary assign-
ment problem. Unfortunately, this constraint is essential, since without it,
there would be feasible solutions involving multiple disconnected cycles.
Despite the similarity, the traveling salesman problem is far more dif-
ficult than the assignment problem. Solving problems having a mere few
hundreds of nodes can be very challenging. By contrast, assignment prob-
lems with hundreds of thousands of nodes can be solved in reasonable time
with the presently available methodology.

Actually, we have already described some discrete/combinatorial prob-

lems that fall within the framework of the minimum cost flow problem, such
as shortest path and assignment (cf. Examples 1.1 and 1.2). These prob-
lems require that the arc flows be 0 or 1, but, as mentioned earlier, we can
neglect these 0-1 constraints because it turns out that even if we relax them
and replace them with flow bound intervals [0, 1], we can obtain optimal
flows that are 0 or 1 (for a proof, see Section 5.2 or Exercise 1.34).
On the other hand, once we deviate from the minimum cost flow struc-
ture and we impose additional constraints or use a nonlinear cost function,
the integer character of optimal solutions is lost, and all integer constraints
must be explicitly imposed. This often complicates dramatically the so-
lution process, and in fact it may be practically impossible to obtain an
exactly optimal solution. As we will discuss in Chapter 10, there are sev-
eral approximate solution approaches that are based on simplified versions
of the problem, such as relaxing the integer constraints. These simpli-
fied problems can often be addressed with the efficient minimum cost flow
algorithms that we will develop in Chapters 2-7.

1.3 NETWORK FLOW ALGORITHMS – AN OVERVIEW

This section, which may be skipped without loss of continuity, provides

a broad classification of the various classes of algorithms for linear and
convex network optimization problems. It turns out that these algorithms
rely on just a few basic ideas, so they can be easily grouped in a few
major categories. By contrast, there is a much larger variety of algorithmic
ideas for discrete optimization problems. For this reason, we postpone the
corresponding discussion for Chapter 10.
Network optimization problems typically cannot be solved analyti-
cally. Usually they must be addressed computationally with one of several
available algorithms. One possibility, for linear and convex problems, is to
use a general purpose linear or nonlinear programming algorithm. How-
ever, the network structure can be exploited to speed up the solution by
Sec. 1.3 Network Flow Algorithms – An Overview 21

using either an adaptation of a general purpose algorithm such as the sim-

plex method, or by using a specialized network optimization algorithm. In
practice, network optimization problems can often be solved hundreds and
even thousands of times faster than general linear or convex programs of
comparable dimension.
The algorithms for linear and convex network problems that we will
discuss in this book can be grouped in three main categories:
(a) Primal cost improvement. Here we try to iteratively improve the
cost to its optimal value by constructing a corresponding sequence of
feasible flows.
(b) Dual cost improvement. Here we define a problem related to the orig-
inal network flow problem, called the dual problem, whose variables
are called prices. We then try to iteratively improve the dual cost to
its optimal value by constructing a corresponding sequence of prices.
Dual cost improvement algorithms also iterate on flows, which are
related to the prices through a property called complementary slack-
ness.
(c) Auction. Here we generate a sequence of prices in a way that is rem-
iniscent of real-life auctions. Strictly speaking, there is no primal or
dual cost improvement here, although we will show that auction can
be viewed as an approximate dual cost improvement process. In ad-
dition to prices, auction algorithms also iterate on flows, which are
related to prices through a property called
-complementary slack-
ness; this is an approximate form of the complementary slackness
property mentioned above.
All of the preceding types of algorithms can be used to solve both
linear and convex network problems (although the structure of the given
problem may favor significantly the use of some types of methods over
others). For simplicity, in this chapter we will explain these ideas primarily
through the assignment problem, deferring a more detailed development to
subsequent chapters. Our illustrations, however, are relevant to the general
minimum cost flow problem and to its convex cost extensions. Some of our
explanations are informal. Precise statements of algorithms and results will
be given in subsequent chapters.

1.3.1 Primal Cost Improvement

Primal cost improvement algorithms for the minimum cost flow problem
start from an initial feasible flow vector and then generate a sequence of
feasible flow vectors, each having a better cost than the preceding one.
Let us derive an important characterization of the differences between suc-
cessive vectors, which is the basis for algorithms as well as for optimality
conditions.
22 Introduction Chap. 1

Let x and x be two feasible flow vectors, and consider their difference
z = x − x . This difference must be a circulation with components
zij = xij − xij ,
since both x and x are feasible. Furthermore, if the cost of x is smaller
than the cost of x, the circulation z must have negative cost, i.e.,

aij zij < 0.
(i,j)∈A

We can decompose z into the sum of simple cycle flows by using the confor-
mal realization theorem (Prop. 1.1). In particular, for some positive integer
K, we have
K
z= wk ξ k ,
k=1
where wk are positive scalars, and ξ k are simple cycle flows whose nonzero
components ξij k are 1 or -1, depending on whether z
ij > 0 or zij < 0,
respectively. It is seen that the cost of z is
 
K
aij zij = wk ck ,
(i,j)∈A k=1

where ck is the cost of the simple cycle flow ξ k . Thus, since the scalars wk
are positive, if the cost of z is negative, at least one ck must be negative.
Note that if Ck is the cycle corresponding to ξ k , we have
  
ck = k =
aij ξij aij − aij ,
(i,j)∈A (i,j)∈Ck+ (i,j)∈Ck−

where Ck+ and Ck− are the sets of forward and backward arcs of the cycle
Ck , respectively. We refer to the expression in the right-hand side above
as the cost of the cycle Ck .
The preceding argument has shown that if x is feasible but not opti-
mal, and x is feasible and has smaller cost than x, then at least one of the
cycles corresponding to a conformal decomposition of the circulation x − x
as above has negative cost. This is used to prove the following important
optimality condition.

Proposition 1.2: Consider the minimum cost flow problem. A flow

vector x∗ is optimal if and only if x∗ is feasible and every simple cycle
C that is unblocked with respect to x∗ has nonnegative cost; that is,
 
aij − aij ≥ 0.
(i,j)∈C + (i,j)∈C −
Sec. 1.3 Network Flow Algorithms – An Overview 23

Proof: Let x∗ be an optimal flow vector and let C be a simple cycle that
is unblocked with respect to x∗ . Then there exists an
> 0 such that
increasing (decreasing) the flow of arcs of C + (of C − , respectively) by

results in a feasible flow that has cost equal to the cost of x∗ plus
times
the cost of C. Thus, since x∗ is optimal, the cost of C must be nonnegative.
Conversely, suppose, to arrive at a contradiction, that x∗ is feasible
and has the nonnegative cycle property stated in the proposition, but is not
optimal. Let x be a feasible flow vector with cost smaller that the one of
x∗ , and consider a conformal decomposition of the circulation z = x − x∗ .
From the discussion preceding the proposition, we see that there is a simple
cycle C with negative cost, such that x∗ij < xij for all (i, j) ∈ C + , and such
that x∗ij > xij for all (i, j) ∈ C − . Since x is feasible, we have bij ≤ xij ≤ cij
for all (i, j). It follows that x∗ij < cij for all (i, j) ∈ C + , and x∗ij > bij for
all (i, j) ∈ C − , so that C is unblocked with respect to x∗ . This contradicts
the hypothesis that every simple cycle that is unblocked with respect to x∗
has nonnegative cost. Q.E.D.

Most primal cost improvement algorithms (including for example the

simplex method, to be discussed in Chapter 5) are based on the preceding
proposition. They employ various mechanisms to construct negative cost
cycles along which flow is pushed without violating the bound constraints.
The idea of improving the cost by pushing flow along a suitable cycle often
has an intuitive meaning as we illustrate in the context of the assignment
problem.

Example 1.7. Multi-Person Exchanges in Assignment

Consider the n × n assignment problem (cf. Example 1.2) and suppose that
we have a feasible assignment, that is, a set of n pairs (i, j) involving each
person i exactly once and each object j exactly once. In order to improve
this assignment, we could consider a two-person exchange, that is, replacing
two pairs (i1 , j1 ) and (i2 , j2 ) from the assignment with the pairs (i1 , j2 ) and
(i2 , j1 ). The resulting assignment will still be feasible, and it will have a
higher value if and only if

ai1 j2 + ai2 j1 > ai1 j1 + ai2 j2 .

We note here that, in the context of the minimum cost flow representation of
the assignment problem, a two-person exchange can be identified with a cycle
involving the four arcs (i1 , j1 ), (i2 , j2 ), (i1 , j2 ), and (i2 , j1 ). Furthermore, this
cycle is the difference between the assignment before and the assignment after
the exchange, while the preceding inequality is equivalent to the cycle having
a positive value.
Unfortunately, it may be impossible to improve the current assignment
by a two-person exchange, even if the assignment is not optimal; see Fig.
1.8. An improvement, however, is possible by means of a k-person exchange,
for some k ≥ 2, where a set of pairs (i1 , j1 ), . . . , (ik , jk ) from the current as-
signment is replaced by the pairs (i1 , j2 ), . . . , (ik−1 , jk ), (ik , j1 ). To see this,
24 Introduction Chap. 1

Value = 1 Figure 1.8: An example of a nonoptimal

1 1
1 11 feasible assignment that cannot be improved
2 by a two-person exchange. The value of
0
each pair is shown next to the correspond-
0
1 1 1 ing arc. Here, the value of the assignment
2 2
{(1, 1), (2, 2), (3, 3)} is left unchanged at 3
2
by any two-person exchange. Through a
2
0
three-person exchange, however, we obtain
1 1
3 3 the optimal assignment, {(1, 2), (2, 3), (3, 1)},
1 which has value 6.

Figure 1.9: Illustration of the correspon-

1 11 dence of a k-person exchange to a simple
cycle. This is the same example as in the
preceding figure. The backward arcs of the
cycle are (1, 1), (2, 2), and (3, 3), and corre-
spond to the current assignment pairs. The
2 2 forward arcs of the cycle are (1, 2), (2, 3),
and (3, 1), and correspond to the new as-
signment pairs. This three-person exchange
is value-improving because the sum of the
values of the forward arcs (2 + 2 + 2) is
3 3 greater than the sum of the values of the
backward arcs (1 + 1 + 1).

note that in the context of the minimum cost flow representation of the as-
signment problem, a k-person exchange corresponds to a simple cycle with
k forward arcs (corresponding to the new assignment pairs) and k backward
arcs (corresponding to the current assignment pairs that are being replaced);
see Fig. 1.9. Thus, performing a k-person exchange is equivalent to pushing
one unit of flow along the corresponding simple cycle. The k-person exchange
improves the assignment if and only if


k−1

k

aik j1 + aim jm+1 − aim jm ,

m=1 m=1

which is equivalent to the corresponding cycle having positive value. Further-

more, by Prop. 1.2, a cost improving cycle exists if the flow corresponding to
the current assignment is not optimal.

1.3.2 Dual Cost Improvement

Duality theory deals with the relation between the original network opti-
mization problem and another optimization problem called the dual . To
develop an intuitive understanding of duality, we will focus on an n × n as-
signment problem (cf. Example 1.2) and consider a closely related economic
equilibrium problem. In particular, let us consider matching the n objects
Sec. 1.3 Network Flow Algorithms – An Overview 25

with the n persons through a market mechanism, viewing each person as

an economic agent acting in his/her own best interest. Suppose that object
j has a price pj and that the person who receives the object must pay the
price pj . Then the net value of object j for person i is aij − pj , and each
person i will logically want to be assigned to an object ji with maximal
value, that is, with

aiji − pji = max {aij − pj }, (1.10)

j∈A(i)

where


A(i) = j | (i, j) ∈ A

is the set of objects that can be assigned to person i. When this condition
holds for all persons i, we say that the assignment and the price vector
p = (p1 , . . . , pn ) satisfy complementary slackness (CS for short); this name
is standard in linear programming. The economic system is then at equi-
librium, in the sense that no person would have an incentive to unilaterally
seek another object. Such equilibrium conditions are naturally of great
interest to economists, but there is also a fundamental relation with the
assignment problem. We have the following proposition.

Proposition 1.3: If a feasible assignment and a set of prices satisfy

the complementary slackness condition (1.10) for all persons i, then
the assignment is optimal and the prices are an optimal solution of
a dual problem, which is to minimize over p = (p1 , . . . , pn ) the cost
function
n 
n
qi (p) + pj ,
i=1 j=1

where the functions qi are given by


qi (p) = max aij − pj , i = 1, . . . , n.
j∈A(i)

Furthermore, the value of the optimal assignment and the optimal cost
of the dual problem are equal.

Proof: The total value of any feasible assignment {(i, ki ) | i = 1, . . . , n}

satisfies
 n n

  n
aiki ≤ max aij − pj + pj , (1.11)
j∈A(i)
i=1 i=1 j=1
26 Introduction Chap. 1

for any set of prices {pj | j = 1, . . . , n}, since the first term of the right-hand
side is no less than
 n
(aiki − pki ) ,
i=1
n
while the second term is equal to i=1 pki . On the other hand, the given
assignment and set of prices, denoted by {(i, ji ) | i = 1, . . . , n} and {pj |
j = 1, . . . , n}, respectively, satisfy the CS conditions, so we have

aiji − pji = max {aij − pj }, i = 1, . . . , n.

j∈A(i)

By adding this relation over all i, we have

n 
  

 n
max aij − pj + pji = aiji
j∈A(i)
i=1 i=1

and by using Eq. (1.11), we obtain

 n 
 
n


aiki ≤ max aij − pj + pji
j∈A(i)
i=1 i=1
n
= aiji
i=1
n

 n
≤ max aij − pj + pj ,
j∈A(i)
i=1 j=1

for every feasible assignment {(i, ki ) | i = 1, . . . , n} and every set of prices

{pj | j = 1, . . . , n}. Therefore, the assignment {(i, ji ) | i = 1, . . . , n} is
optimal for the primal problem, and the set of prices {pj | j = 1, . . . , n}
is optimal for the dual problem. Furthermore, the two optimal values are
equal. Q.E.D.

In analogy with primal cost improvement algorithms, one may start

with a price vector and try to successively obtain new price vectors with
improved dual cost. The major algorithms of this type involve price changes
of the form 
pi + γ if i ∈ S,
pi := (1.12)
pi if i ∈
/ S,
where S is a connected subset of nodes, and γ is some positive scalar that
is small enough to ensure that the new price vector has an improved dual
cost.
The existence of a node subset S that results in cost improvement at
a nonoptimal price vector, as described above, will be shown in Chapter 6.
Sec. 1.3 Network Flow Algorithms – An Overview 27

This is an important and remarkable result, which may be viewed as a dual

version of the result of Prop. 1.2 (at a nonoptimal flow vector, there exists
at least one unblocked simple cycle with negative cost). In fact both results
are special cases of a more general theorem concerning elementary vectors
of subspaces, which is central in the theory of monotropic programming
(see Chapter 9).
Most dual cost improvement methods, simultaneously with changing
p along a direction of dual cost improvement, also iterate on a flow vector
x satisfying CS together with p. They terminate when x becomes feasible,
at which time, by Prop. 1.3, the pair (x, p) must consist of a primal and a
dual optimal solution.
In Chapter 6 we will discuss two main methods that select subsets S
and corresponding directions of dual cost improvement in different ways:
(a) In the primal-dual method , the direction has a steepest ascent prop-
erty, that is, it provides the maximal rate of improvement of the dual
cost per unit change in the price vector.
(b) In the relaxation (or coordinate ascent) method , the direction is com-
puted so that it has a small number of nonzero elements (i.e., the set
S has few nodes). Such a direction may not be optimal in terms of
rate of dual cost improvement, but can typically be computed much
faster than the steepest ascent direction. Often the direction has only
one nonzero element, in which case only one node price coordinate is
changed; this motivates the name “coordinate ascent.” Note, how-
ever, that coordinate ascent directions cannot be used exclusively to
improve the dual cost, as is shown in Fig. 1.10.

1.3.3 Auction

Our third type of algorithm represents a significant departure from the

cost improvement idea; at any one iteration, it may deteriorate both the
primal and the dual cost, although in the end it does find an optimal primal
solution. It is based on an approximate version of complementary slackness,
called
-complementary slackness, and while it implicitly tries to solve a
dual problem, it actually attains a dual solution that is not quite optimal.
This subsection introduces the main ideas underlying auction algorithms.
Chapters 7 and 9 provide a detailed discussion for the minimum cost flow
problem and for the separable convex cost problem, respectively.

Naive Auction

Let us return to the assignment problem, and consider a natural process

for finding an equilibrium assignment and price vector. We will call this
process the naive auction algorithm, because it has a serious flaw, as will be
28 Introduction Chap. 1

p
2

Surfaces of Equal
Dual Cost

Figure 1.10: (a) The difficulty with

using exclusively coordinate ascent it-
erations to solve the dual problem.
p1 Because the dual cost is piecewise lin-
(a) ear, it may be impossible to improve
it at some corner points by chang-
p2 ing any single price coordinate. (b)
As will be discussed in Chapter 6, a
Surfaces of Equal dual cost improvement is possible by
Dual Cost changing several price coordinates by
equal amounts, as in Eq. (1.12).

p1
(b)

seen shortly. Nonetheless, this flaw will help motivate a more sophisticated
and correct algorithm.
The naive auction algorithm proceeds in iterations and generates a
sequence of price vectors and partial assignments. By a partial assignment
we mean an assignment where only a subset of the persons have been
matched with objects. A partial assignment should be contrasted with a
feasible or complete assignment where all the persons have been matched
with objects on a one-to-one basis. At the beginning of each iteration, the
CS condition [cf. Eq. (1.10)]

aiji − pji = max {aij − pj }

j∈A(i)

is satisfied for all pairs (i, ji ) of the partial assignment. If all persons
are assigned, the algorithm terminates. Otherwise some person who is
unassigned, say i, is selected. This person finds an object ji which offers
maximal value, that is,

ji = arg max {aij − pj },

j∈A(i)

and then:
(a) Gets assigned to the best object ji ; the person who was assigned to
ji at the beginning of the iteration (if any) becomes unassigned.
Sec. 1.3 Network Flow Algorithms – An Overview 29

(b) Sets the price of ji to the level at which he/she is indifferent between
ji and the second best object; that is, he/she sets pji to

pji + γi ,

where
γi = vi − wi , (1.13)
vi is the best object value,

vi = max {aij − pj }, (1.14)

j∈A(i)

and wi is the second best object value,

wi = max {aij − pj }. (1.15)

j∈A(i), j=ji

(Note that as pji is increased, the value aiji − pji offered by object ji
to person i is decreased. γi is the largest increment by which pji can
be increased, while maintaining the property that ji offers maximal
value to i.)
This process is repeated in a sequence of iterations until each person has
been assigned to an object.
We may view this process as an auction where at each iteration the
bidder i raises the price of a preferred object by the bidding increment γi .
Note that γi cannot be negative, since vi ≥ wi [compare Eqs. (1.14)and
(1.15)], so the object prices tend to increase. The choice γi is illustrated
in Fig. 1.11. Just as in a real auction, bidding increments and price in-
creases spur competition by making the bidder’s own preferred object less
attractive to other potential bidders.

-Complementary Slackness

Unfortunately, the naive auction algorithm does not always work (although
it is an excellent initialization procedure for other methods, such as primal-
dual or relaxation, and it is useful in other specialized contexts). The diffi-
culty is that the bidding increment γi is 0 when two or more objects are tied
in offering maximum value for the bidder i. As a result, a situation may be
created where several persons contest a smaller number of equally desirable
objects without raising their prices, thereby creating a never ending cycle;
see Fig. 1.12.
To break such cycles, we introduce a perturbation mechanism, moti-
vated by real auctions where each bid for an object must raise its price by
a minimum positive increment, and bidders must on occasion take risks to
win their preferred objects. In particular, let us fix a positive scalar
, and
30 Introduction Chap. 1

v : The value of j , the best object for person i

i i
-------------
Bidding increment γi of person i for its best
object ji
-------------
w i : The value of the second best object for person i

Values a ij - p j
of objects j
for person i

Figure 1.11: In the naive auction algorithm, even after the price of the best
object ji is increased by the bidding increment γi , ji continues to be the best
object for the bidder i, so CS is satisfied at the end of the iteration. However, we
have γi = 0 if there is a tie between two or more objects that are most preferred
by i.

say that a partial assignment and a price vector p satisfy
-complementary

slackness (
-CS for short) if

aij − pj ≥ max {aik − pk } −

k∈A(i)

for all assigned pairs (i, j). In words, to satisfy
-CS, all assigned persons
of the partial assignment must be assigned to objects that are within
of
being best.

The Auction Algorithm

We now reformulate the previous auction process so that the bidding in-
crement is always at least equal to
. The resulting method, the auction
algorithm, is the same as the naive auction algorithm, except that the
bidding increment γi is
γi = vi − wi +
(1.16)
rather than γi = vi − wi as in Eq. (1.13). With this choice, the
-CS
condition is satisfied, as illustrated in Fig. 1.13. The particular increment
γi = vi − wi +
used in the auction algorithm is the maximum amount with
this property. Smaller increments γi would also work as long as γi ≥
,
but using the largest possible increment accelerates the algorithm. This
is consistent with experience from real auctions, which tend to terminate
faster when the bidding is aggressive.
Sec. 1.3 Network Flow Algorithms – An Overview 31

PERSONS OBJECTS

Initially assigned 1 1 Initial price = 0

to object 1

Initially assigned
2 2 Initial price = 0
to object 2
Here a ij= C > 0 for all (i,j) with i = 1,2,3 and j = 1,2
and a = 0 for all (i,j) with i = 1,2,3 and j = 3
ij
Initially 3 3
unassigned Initial price = 0

At Start of Object Assigned Bidder Preferred Bidding

Iteration # Prices Pairs Object Increment

1 0,0,0 (1,1), (2,2) 3 2 0

2 0,0,0 (1,1), (3,2) 2 2 0
3 0,0,0 (1,1), (2,2) 3 2 0

Figure 1.12: Illustration of how the naive auction algorithm may never terminate
for a problem involving three persons and three objects. Here objects 1 and 2
offer benefit C > 0 to all persons, and object 3 offers benefit 0 to all persons. The
algorithm cycles as persons 2 and 3 alternately bid for object 2 without changing
its price because they prefer equally object 1 and object 2.

It can be shown that this reformulated auction process terminates,

necessarily with a feasible assignment and a set of prices that satisfy
-
CS. To get a sense of this, note that if an object receives a bid during
m iterations, its price must exceed its initial price by at least m
. Thus,
for sufficiently large m, the object will become “expensive” enough to be
judged “inferior” to some object that has not received a bid so far. It follows
that only for a limited number of iterations can an object receive a bid while
some other object still has not yet received any bid. On the other hand,
once every object has received at least one bid, the auction terminates.
(This argument assumes that any person can bid for any object, but it can
be generalized to the case where the set of feasible person-object pairs is
limited, as long as at least one feasible assignment exists; see Prop. 7.2 in
Chapter 7.) Figure 1.14 shows how the auction algorithm, based on the
bidding increment γi = vi − wi +
[see Eq. (1.16)], overcomes the cycling
difficulty in the example of Fig. 1.12.
When the auction algorithm terminates, we have an assignment sat-
isfying
-CS, but is this assignment optimal? The answer depends strongly
32 Introduction Chap. 1

v : The value of j i , the best object for person i

------------- i

Bidding increment γ i of person i for its best

object j
i
-------------
wi : The value of the second best object for person i
ε
------------
Values a ij - pj
of objects j
for person i

Figure 1.13: In the auction algorithm, even after the price of the preferred
object ji is increased by the bidding increment γi , ji will be within  of being
most preferred, so the -CS condition holds at the end of the iteration.

on the size of
. In a real auction, a prudent bidder would not place an

excessively high bid for fear of winning the object at an unnecessarily high
price. Consistent with this intuition, we can show that if
is small, then
the final assignment will be “almost optimal.” In particular, we will show
that the total benefit of the final assignment is within n
of being optimal .
The idea is that a feasible assignment and a set of prices satisfying
-CS
may be viewed as satisfying CS for a slightly different problem, where all
benefits aij are the same as before except the benefits of the n assigned
pairs, which are modified by no more than
.

Proposition 1.4: A feasible assignment satisfying
-complementary

slackness, together with some price vector, attains within n
the opti-
mal primal value. Furthermore, the price vector attains within n
the
optimal dual cost.

Proof: Let A∗ be the optimal total assignment benefit


n
A∗ = max aiki
ki , i=1,...,n
ki =km if i=m i=1

and let D∗ be the optimal dual cost (cf. Prop. 1.3):

⎧ ⎫
⎨ n

 n ⎬
D∗ = min max aij − p j + p j .
pj ⎩ j∈A(i) ⎭
j=1,...,n i=1 j=1
Sec. 1.3 Network Flow Algorithms – An Overview 33

PERSONS OBJECTS

Initially assigned 1 1 Initial price = 0

to object 1

Initially assigned
to object 2
2 2 Initial price = 0

Here a ij= C > 0 for all (i,j) with i = 1,2,3 and j = 1,2
and a ij = 0 for all (i,j) with i = 1,2,3 and j = 3
Initially 3 3 Initial price = 0
unassigned

At Start of Object Assigned Bidder Preferred Bidding

Iteration # Prices Pairs Object Increment

1 0,0,0 (1,1), (2,2) 3 2 

2 0,,0 (1,1), (3,2) 2 1 2
3 2,,0 (2,1), (3,2) 1 2 2
4 2,3,0 (1,2), (2,1) 3 1 2
5 4,3,0 (1,2), (3,1) 2 2 2
6 ··· ··· ··· ··· ···

Figure 1.14: Illustration of how the auction algorithm, by making the bidding
increment at least , overcomes the cycling difficulty for the example of Fig. 1.12.
The table shows one possible sequence of bids and assignments generated by
the auction algorithm, starting with all prices equal to 0 and with the partial
assignment {(1, 1), (2, 2)}. At each iteration except the last, the person assigned
to object 3 bids for either object 1 or 2, increasing its price by  in the first iteration
and by 2 in each subsequent iteration. In the last iteration, after the prices of 1
and 2 reach or exceed C, object 3 receives a bid and the auction terminates.

If {(i, ji ) | i = 1, . . . , n} is the given assignment satisfying the
-CS condi-

tion together with a price vector p, we have
max {aij − pj } −
≤ aiji − pji .
j∈A(i)

By adding this relation over all i, we see that

n   

 n
D∗ ≤ max aij − pj + pji ≤ aiji + n
≤ A∗ + n
.
j∈A(i)
i=1 i=1

Since we showed in Prop. 1.3 that A∗

= D∗ , it follows that the total
n
assignment benefit i=1 aiji is within n
of the optimal value A∗ , while
the dual cost of p is within n
of the optimal dual cost. Q.E.D.
34 Introduction Chap. 1

Suppose now that the benefits aij are all integer, which is the typical
practical case. (If aij are rational numbers, they can be scaled up to integer
by multiplication with a suitable common number.) Then the total benefit
of any assignment is integer, so if n
< 1, any complete assignment that is
within n
of being optimal must be optimal. It follows that if
1

<
n
and the benefits aij are all integer, then the assignment obtained upon ter-
mination of the auction algorithm is optimal .
Figure 1.15 shows the sequence of generated object prices for the ex-
ample of Fig. 1.12 in relation to the contours of the dual cost function.
It can be seen from this figure that each bid has the effect of setting the
price of the object receiving the bid nearly equal (within
) to the price
that minimizes the dual cost with respect to that price, with all other
prices held fixed (this will be shown rigorously in Section 7.1). Successive
minimization of a cost function along single coordinates is a central fea-
ture of coordinate descent and relaxation methods, which are popular for
unconstrained minimization of smooth functions and for solving systems
of smooth equations. Thus, the auction algorithm can be interpreted as
an approximate coordinate descent method; as such, it is related to the
relaxation method discussed in the previous subsection.

Scaling

Figure 1.15 also illustrates a generic feature of auction algorithms. The

amount of work needed to solve the problem can depend strongly on the
value of
and on the maximum absolute object benefit

C = max |aij |.
(i,j)∈A

Basically, for many types of problems, the number of iterations up to termi-

nation tends to be proportional to C/
. This can be seen from the figure,
where the total number of iterations is roughly C/
, starting from zero
initial prices.
Note also that there is a dependence on the initial prices; if these
prices are “near optimal,” we expect that the number of iterations needed
to solve the problem will be relatively small. This can be seen from Fig.
1.15; if the initial prices satisfy p1 ≈ p3 + C and p2 ≈ p3 + C, the number
of iterations up to termination is quite small.
The preceding observations suggest the idea of
-scaling, which con-
sists of applying the algorithm several times, starting with a large value of

and successively reducing
until it is less than some critical value (for
example, 1/n, when aij are integer). Each application of the algorithm pro-
vides good initial prices for the next application. This is a common idea
Sec. 1.3 Network Flow Algorithms – An Overview 35

p
2

Contours of the
dual function

C
ε
ε

ε
ε

Price p 3 is fixed at 0
ε
ε
C p1

Figure 1.15: A sequence of prices p1 and p2 generated by the auction algorithm

for the example of Figs. 1.12 and 1.14. The figure shows the equal dual cost
surfaces in the space of p1 and p2 , with p3 fixed at 0. The arrows indicate the
price iterates as given by the table of Fig. 1.14. Termination occurs when the prices
reach an -neighborhood of the point (C, C), and object 3 becomes “sufficiently
inexpensive” to receive a bid and to get assigned. The total number of iterations
is roughly C/, starting from zero initial prices.

in nonlinear programming; it is encountered, for example, in barrier and

penalty function methods (see Section 8.8). In practice, scaling is typically
beneficial, and accelerates the termination of the auction algorithm.

1.3.4 Good, Bad, and Polynomial Algorithms

We have discussed several types of methods, so the natural question arises:

is there a best method and what criterion should we use to rank methods?
A practitioner who has a specific type of problem to solve, perhaps
repeatedly, with the data and size of the problem within some limited range,
will usually be interested in one or more of the following:
(a) Fast solution time.
(b) Flexibility to use good starting solutions (which the practitioner can
usually provide, based on his/her knowledge of the problem, or based
on a known solution of some similar problem).
36 Introduction Chap. 1

(c) The ability to perform sensitivity analysis (resolve the problem with
slightly different problem data) quickly.
(d) The ability to take advantage of parallel computing hardware.
Given the diversity of these considerations, it is not surprising that
there is no algorithm that will dominate the others in all or even most
practical situations. Otherwise expressed, every type of algorithm that we
will discuss is best given the right type of practical situation. Thus, to
make intelligent choices, the practitioner needs to understand the proper-
ties of different algorithms relating to speed of convergence, flexibility, par-
allelization, and suitability for specific problem structures. For challenging
problems, the choice of algorithm is often settled by experimentation with
several candidates.
A theoretical analyst may also have difficulty ranking different algo-
rithms for specific types of problems. The most common approach for this
purpose is worst-case computational complexity analysis. For example, for
the minimum cost flow problem, one tries to bound the number of elemen-
tary numerical operations needed by a given algorithm with some measure
of the “problem size,” that is, with some expression of the form
Kf (N, A, C, U, S),
where
N is the number of nodes,
A is the number of arcs,
C is the arc cost range max(i,j)∈A |aij |,
U is the maximum arc flow range max(i,j)∈A (cij − bij ),
S is the supply range maxi∈N |si |,
f is some known function,
K is a (usually unknown) constant.
If a bound of this form can be found,  we say that the running time or
operation count of the algorithm is O f (N, A, C, U, S) . If f (N, A, C, U, S)
can be written as a polynomial function of the number of bits needed to
express the problem data, the algorithm is said
 to be polynomial
 . Exam- 
ples of polynomial complexity bounds are O N α Aβ and O N α Aβ log C ,
where α and β are positive
 integers,
 and the numbers aij are assumed in-
teger. The bound O N α Aβ is sometimes said to be strongly polynomial
because
 it involves
 only the graph size parameters. A bound of the form
O N α Aβ C is not polynomial, even assuming that the aij are integer, be-
cause C is not a polynomial expression of log C, the number
 of bits needed
to express a single number aij . Bounds like O N α Aβ C , which are poly-
nomial in the problem data rather than in the number of bits needed to
express the data, are called pseudopolynomial .
 Sec. 1.4 Notes, Sources, and Exercises 37

A common assumption in theoretical computer science is that poly-

nomial algorithms are “better” than pseudopolynomial, and pseudopoly-
nomial algorithms are “better” than exponential [for example, those with
a bound of the form K2g(N,A) , where g is a polynomial in N and A]. Fur-
thermore, it is thought that two polynomial algorithms can be compared in
terms of the degree of the polynomial bound; e.g., an O(N 2 ) algorithm is
“better” than an O(N 3 ) algorithm. Unfortunately, quite often this assump-
tion is not supported by computational practice in linear programming and
network optimization. Pseudopolynomial and even exponential algorithms
are often faster in practice than polynomial ones. In fact, the simplex
method for general linear programs is an exponential algorithm, as shown
by Klee and Minty [1972] (see also the textbooks by Chvatal [1983], or
Bertsimas and Tsitsiklis [1997]), and yet it is used widely, because of its
excellent practical properties.
There are two main reasons why worst-case complexity estimates may
fail to predict the practical performance of network flow algorithms. First,
the estimates, even if they are tight, may be very pessimistic as they may
correspond to problem instances that are highly unlikely in practice. (Av-
erage complexity estimates would be more appropriate for such situations.
However, obtaining these is usually hard, and the statistical assumptions
underlying them may be inappropriate for many types of practical prob-
lems.) Second, worst-case complexity estimates involve the (usually un-
known) constant K, which may dominate the estimate for all except for
unrealistically large problem sizes. Thus, a comparison between two algo-
rithms that is based on the size-dependent terms of running time estimates,
and does not take into account the corresponding constants may be unre-
liable.
Despite its shortcomings, computational complexity analysis is valu-
able because it often illuminates the computational bottlenecks of many al-
gorithms and motivates the use of efficient data structures. For this reason,
throughout the book, we will comment on available complexity results, we
will prove some of the most important estimates, and we will try to relate
these estimates to computational practice. For some classes of problems,
however, it turns out that the methods with the best computational com-
plexity are impractical, because they are either too complicated or too slow
in practice. In such cases, we will refer to the literature, without providing
a detailed discussion.

1.4 NOTES, SOURCES, AND EXERCISES

Network problems are discussed in many books (Berge [1962], Berge and
Ghouila-Houri [1962], Ford and Fulkerson [1962], Dantzig [1963], Busacker
and Saaty [1965], Hu [1969], Iri [1969], Frank and Frisch 1970], Christofides
38 Introduction Chap. 1

[1975], Zoutendijk [1976], Minieka [1978], Jensen and Barnes [1980], Ken-
nington and Helgason [1980], Papadimitriou and Steiglitz [1982], Chvatal
[1983], Gondran and Minoux [1984], Luenberger [1984], Rockafellar [1984],
Bazaraa, Jarvis, and Sherali [1990], Bertsekas [1991a], Murty [1992], Bert-
simas and Tsitsiklis [1997]). Several of these books discuss linear program-
ming first and develop linear network optimization as a special case. An
alternative approach that relies heavily on duality, is given by Rockafellar
[1984]. The conformal realization theorem (Prop. 1.1) has been developed
in different forms in several sources, including Ford and Fulkerson [1962],
Busacker and Saaty [1965], and Rockafellar [1984].
The primal cost improvement approach for network optimization was
initiated by Dantzig [1951], who specialized the simplex method to the
transportation problem. The extensive subsequent work using this ap-
proach is surveyed at the end of Chapter 5.
The dual cost improvement approach was initiated by Kuhn [1955]
who proposed the Hungarian method for the assignment problem. (The
name of the algorithm honors its connection with the research of the Hun-
garian mathematicians Egervary [1931] and König [1931].) Work using this
approach is surveyed in Chapter 6.
The auction approach was initiated in Bertsekas [1979a] for the as-
signment problem, and in Bertsekas [1986a], [1986b] for the minimum cost
flow problem. Work using this approach is surveyed at the end of Chapter
7.

EXERCISES

1.1

Consider the graph and the flow vector of Fig. 1.16.

(a) Enumerate the simple paths and the simple forward paths that start at
node 1.
(b) Enumerate the simple cycles and the simple forward cycles of the graph.
(c) Is the graph connected? Is it strongly connected?
(d) Calculate the divergences of all the nodes and verify that they add to 0.
(e) Give an example of a simple path flow that starts at node 1, ends at node
5, involves four arcs, and conforms to the given flow vector.
(f) Suppose that all arcs have arc flow bounds -1 and 5. Enumerate all the
simple paths that start at node 1, end at node 5, and are unblocked with
Sec. 1.4 Notes, Sources, and Exercises 39

respect to the given flow vector.

2 5
1
3 Figure 1.16: Flow vector for Ex-
1 2 -1 4 5 ercise 1.1. The arc flows are the
2
-1 numbers shown next to the arcs.
2 1
3

1.2 (Proof of the Conformal Realization Theorem)

Prove the conformal realization theorem (Prop. 1.1) by completing the details
of the following argument. Assume first that x is a circulation. Consider the
following procedure by which given x, we obtain a simple cycle flow x that
conforms to x and satisfies

0 ≤ xij ≤ xij for all arcs (i, j) with 0 ≤ xij ,

xij ≤ xij ≤0 for all arcs (i, j) with xij ≤ 0,
xij = xij for at least one arc (i, j) with xij = 0;

(see Fig. 1.17). Choose an arc (i, j) with xij = 0. Assume that xij > 0. (A
similar procedure can be used when xij < 0.) Construct a sequence of node
subsets T0 , T1 , . . ., as follows: Take T0 = {j}. For k = 0, 1, . . ., given Tk , let


/ ∪kp=0 Tp | there is a node m ∈ Tk , and either an arc (m, n)
Tk+1 = n ∈

such that xmn > 0 or an arc (n, m) such that xnm < 0 ,

and mark each node n ∈ Tk+1 with the label “(m, n)” or “(n, m),” where m
is a node of Tk such that xmn > 0 or xnm < 0, respectively. The procedure
terminates when Tk+1 is empty.
At the end of the procedure, trace labels backward from i until node j is
reached. (How do we know that i belongs to one of the sets Tk ?) In particular,
let “(i1 , i)” or “(i, i1 )” be the label of i, let “(i2 , i1 )” or “(i1 , i2 )” be the label
of i1 , etc., until a node ik with label “(ik , j)” or “(j, ik )” is found. The cycle
C = (j, ik , ik−1 , . . . , i1 , i, j) is simple, it contains (i, j) as a forward arc, and is
such that all its forward arcs have positive flow and all its backward arcs have
negative flow. Let a = min(m,n)∈C |xmn | > 0. Then the simple cycle flow x ,
where 
a if (i, j) ∈ C + ,

xij = −a if (i, j) ∈ C − ,
0 otherwise,
has the required properties.
Now subtract x from x. We have xij − xij > 0 only for arcs (i, j) with
xij > 0, xij − xij < 0 only for arcs (i, j) with xij < 0, and xij − xij = 0 for at
40 Introduction Chap. 1

Flow < 0 Flow > 0

T0
m
j Flow < 0
T1
Flow > 0
n
ik
x ij > 0 T2
Flow > 0 i k-1
C
i
Flow < 0
Flow > 0 Tk
i k-2
i1
Flow > 0 i2

Figure 1.17: Construction of a cycle of arcs with nonzero flow used in the proof
of the conformal realization theorem.

least one arc (i, j) with xij = 0. If x is integer, then x and x − x will also be
integer. We then repeat the process (for at most A times) with the circulation x
replaced by the circulation x − x and so on, until the zero flow is obtained.
If x is not a circulation, we form an enlarged graph by introducing a new
node s and by introducing for each node i ∈ N an arc (s, i) with flow xsi equal
to the divergence yi . The resulting flow vector is seen to be a circulation in the
enlarged graph (why?). This circulation, by the result just shown, can be decom-
posed into at most A + N simple cycle flows of the enlarged graph, conforming
to the flow vector. Out of these cycle flows, we consider those containing node
s, and we remove s and its two incident arcs while leaving the other cycle flows
unchanged. As a result we obtain a set of at most A+N path flows of the original
graph, which add up to x. These path flows also conform to x, as required.

1.3

Use the algorithm of Exercise 1.2 to decompose the flow vector of Fig. 1.16 into
conforming simple path flows.

1.4 (Path Decomposition Theorem)

(a) Use the conformal realization theorem (Prop. 1.1) to show that a forward
path P can be decomposed into a (possibly empty) collection of simple
forward cycles, together with a simple forward path that has the same
start node and end node as P . (Here “decomposition” means that the
Sec. 1.4 Notes, Sources, and Exercises 41

union of the arcs of the component paths is equal to the set of arcs of P
with the multiplicity of repeated arcs properly accounted for.)
(b) Suppose that a graph is strongly connected and that a length aij is given for
every arc (i, j). Show that if all forward cycles have nonnegative length,
then there exists a shortest path from any node s to any node t. Show
also that if there exists a shortest path from some node s to some node t,
then all forward cycles have nonnegative length. Why is the connectivity
assumption needed?

1.5 (Cycle Decomposition - Euler Cycles)

Consider a graph such that each of the nodes has even degree.
(a) Give an algorithm to decompose the graph into a collection of simple cycles
that are disjoint, in the sense that they share no arcs (although they may
share some nodes). (Here “decomposition” means that the union of the
arcs of the component cycles is equal to the set of arcs of the graph.) Hint:
Given a connected graph where each of the nodes has even degree, the
deletion of the arcs of any cycle creates some connected subgraphs where
each of the nodes has even degree (including possibly some isolated nodes).
(b) Assume in addition that the graph is connected. Show that there is an
Euler cycle, i.e., a cycle that contains all the arcs of a graph exactly once.
Hint: Apply the decomposition of part (a), and successively merge an Euler
cycle of a subgraph with a simple cycle.

1.6

In the graph of Fig. 1.16, consider the graph obtained by deleting node 1 and
arcs (1, 2), (1, 3), and (5, 4). Decompose this graph into a collection of simple
cycles that are disjoint (cf. Exercise 1.5) and construct an Euler cycle.

1.7

(a) Consider an n × n chessboard, and a rook that is allowed to make the

standard moves along the rows and columns. Show that the rook can start
at a given square and return to that square after making each of the possible
legal moves exactly once and in one direction only [of the two moves (a, b)
and (b, a) only one should be made]. Hint: Construct an Euler cycle in a
suitable graph.
(b) Consider an n × n chessboard with n even, and a bishop that is allowed to
make two types of moves: legal moves (which are the standard moves along
the diagonals of its color), and illegal moves (which go from any square of
its color to any other square of its color). Show that the bishop can start at
a given square and return to that square after making each of the possible
legal moves exactly once and in one direction only, plus n2 /4 illegal moves.
42 Introduction Chap. 1

For every square of its color, there should be exactly one illegal move that
either starts or ends at that square.

1.8 (Forward Euler Cycles)

Consider a graph and the question whether there exists a forward cycle that
passes through each arc of the graph exactly once. Show that such a cycle exists
if and only if the graph is connected and the number of incoming arcs to each
node is equal to the number of outgoing arcs from the node.

1.9

Consider an n × n chessboard with n ≥ 4. Show that a knight starting at any

square can visit every other square, with a move sequence that contains every
possible move exactly once [a move (a, b) as well as its reverse (b, a) should be
made]. Interpret this sequence as a forward Euler cycle in a suitable graph (cf.
Exercise 1.8).

1.10 (Euler Paths)

Consider a graph and the question whether there exists a path that passes through
each arc of the graph exactly once. Show that such a path exists if and only if
the graph is connected, and either the degrees of all the nodes are even, or else
the degrees of all the nodes except two are even.

1.11

In shatranj, the old version of chess, the firz (or vizier, the predecessor to the
modern queen) can move one square diagonally in each direction. Show that
starting at a corner of an n × n chessboard where n is even, the firz can reach
the opposite corner after making each of the possible moves along its diagonals
exactly once and in one direction only [of the two moves (a, b) and (b, a) only one
should be made].

1.12

Show that the number of nodes with odd degree in a graph is even.

1.13

Assume that all the nodes of a graph have degree greater than one. Show that
the graph must contain a cycle.
Sec. 1.4 Notes, Sources, and Exercises 43

1.14

(a) Show that every tree with at least two nodes has at least two nodes with
degree one.
(b) Show that a graph is a tree if and only if it is connected and the number
of arcs is one less than the number of nodes.

1.15

Consider a volleyball net that consists of a mesh with m squares on the horizontal
dimension and n squares on the vertical. What is the maximum number of strings
that can be cut before the net falls apart into two pieces.

1.16 (Checking Connectivity)

Consider a graph with A arcs.

(a) Devise an algorithm with O(A) running time that checks whether the graph
is connected, and if it is connected, simultaneously constructs a path con-
necting any two nodes. Hint: Start at a node, mark its neighbors, and
continue.
(b) Repeat part (a) for the case where we want to check strong connectedness.
(c) Devise an algorithm with O(A) running time that checks whether there
exists a cycle that contains two given nodes.
(d) Repeat part (c) for the case where the cycle is required to be forward.

1.17 (Inequality Constrained Minimum Cost Flows)

Consider the following variant of the minimum cost flow problem:


minimize aij xij
(i,j)∈A
 
subject to si ≤ xij − xji ≤ si , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

bij ≤ xij ≤ cij , ∀ (i, j) ∈ A,

where the bounds si and si on the divergence of node i are given. Show that
this problem can be converted to a standard (equality constrained) minimum
cost flow problem by adding an extra node A and an arc (A, i) from this node to
every other node i, with feasible flow range [0, si − si ].
44 Introduction Chap. 1

1.18 (Node Throughput Constraints)

Consider the minimum cost flow problem with the additional constraints that
the total flow of the outgoing arcs from each node i must lie within a given range
[ti , ti ], that is,

ti ≤ xij ≤ ti .
{j|(i,j)∈A}

Convert this problem into the standard form of the minimum cost flow problem
by splitting each node into two nodes with a connecting arc.

1.19 (Piecewise Linear Arc Costs)

Consider the minimum cost flow problem with the difference that, instead of the
linear form aij xij , each arc’s cost function has the piecewise linear form

a1ij xij if bij ≤ xij ≤ mij ,
fij (xij ) =
a1ij mij + a2ij (xij − mij ) if mij ≤ xij ≤ cij ,

where mij , a1ij , and a2ij are given scalars satisfying bij ≤ mij ≤ cij and a1ij ≤ a2ij .
(a) Show that the problem can be converted to a linear minimum cost flow
problem where each arc (i, j) is replaced by two arcs with arc cost coeffi-
cients a1ij and a2ij , and arc flow ranges [bij , mij ] and [0, cij − mij ], respec-
tively.
(b) Generalize to the case of piecewise linear cost functions with more than
two pieces.

1.20 (Asymmetric Assignment and Transportation Problems)

Consider an assignment problem where the number of objects is larger than the
number of persons, and we require that each person be assigned to one object.
The associated linear program (cf. Example 1.2) is

maximize aij xij
(i,j)∈A

subject to xij = 1, ∀ i = 1, . . . , m,
{j|(i,j)∈A}

xij ≤ 1, ∀ j = 1, . . . , n,
{i|(i,j)∈A}

0 ≤ xij ≤ 1, ∀ (i, j) ∈ A,

where m < n.
(a) Show how to formulate this problem as a minimum cost flow problem by
introducing extra arcs and nodes.
Sec. 1.4 Notes, Sources, and Exercises 45

(b) Repeat part (a) for the case where there may be some persons that are
left unassigned; that is, the constraint x = 1 is replaced by
{j|(i,j)∈A} ij

{j|(i,j)∈A}
xij ≤ 1. Give an example of a problem with aij > 0 for all
(i, j) ∈ A, which is such that in the optimal assignment some persons are
left unassigned, even though there exist feasible assignments that assign
every person to some object.
(c) Formulate an asymmetric transportation problem where the total supply
is less than the total demand, but some demand may be left unsatisfied,
and appropriately modify your answers to parts (a) and (b).

1.21 (Bipartite Matching)

Bipartite matching problems are assignment problems where the coefficients (i, j)
are all equal to 1. In such problems, we want to maximize the cardinality of the
assignment, that is, the number of assigned pairs (i, j). Formulate a bipartite
matching problem as an equivalent max-flow problem.

1.22 (Production Planning)

Consider a problem of scheduling production of a certain item to meet a given

demand over N time periods. Let us denote:
xi : The amount of product stored at the beginning of period i, where
i = 0, . . . , N − 1. There is a nonnegativity constraint on xi .
ui : The amount of product produced during period i. There is a constraint
0 ≤ ui ≤ ci , where the scalar ci is given for each i.
di : The amount of product demanded during period i. This is a given
scalar for each i.
The amount of product stored evolves according to the equation
xi+1 = xi + ui − di , i = 0, . . . , N − 1.
Given x0 , we want to find a feasible production sequence {u0 , . . . , uN −1 } that
minimizes

N −1

(ai xi + bi ui ),
i=0

where ai and bi are given scalars for each i. Formulate this problem as a minimum
cost flow problem. Hint: For each i, introduce a node that connects to a special
artificial node.

1.23 (Capacity Expansion)

The capacity of a certain facility is to be expanded over N time periods by adding

an increment ui ∈ [0, ci ] at time period i = 0, . . . , N −1, where ci is a given scalar.
Thus, if xi is the capacity at the beginning of period i, we have
xi+1 = xi + ui , i = 0, . . . , N − 1.
46 Introduction Chap. 1

Given x0 , consider the problem of finding ui , i = 0, . . . , N − 1, such that each xi

lies within a given interval [xi , x̄i ] and the cost


N −1

(ai xi + bi ui )
i=0

is minimized, where ai and bi are given scalars for each i. Formulate the problem
as a minimum cost flow problem.

1.24 (Dynamic Transhipment Problems)

Consider a transhipment context for the minimum cost flow problem where the
problem is to optimally transfer flow from some supply points to some demand
points over arcs of limited capacity. In a dynamic version of this context, the
transfer is to be performed over N time units, and transferring flow along an arc
(i, j) requires time τij , which is a given positive integer number of time units.
This means that at each time t = 0, . . . , N − τij , we may send from node i along
arc (i, j) a flow xij ∈ [0, cij ], which will arrive at node j at time t+τij . Formulate
this problem as a minimum cost flow problem involving a copy of the given graph
for each time period.

1.25 (Concentrator Assignment)

We have m communication terminals, each to be connected to one out of a

given collection of concentrators. Suppose that there is a cost aij for connecting
terminal i to concentrator j, and that each concentrator j has an upper bound
bj on the number of terminals it can be connected to. Also, each terminal i can
be connected to only a given subset of concentrators.
(a) Formulate the problem of finding the minimum cost connection of terminals
to concentrators as a minimum cost flow problem. Hint: You may use the
fact that there exists an integer optimal solution to a minimum cost flow
problem with integer supplies and arc flow bounds. (This will be shown in
Chapter 5.)
(b) Suppose that a concentrator j can operate in an overload condition with
a number of connected terminals greater than bj , up to a number bj > bj .
In this case, however, the cost per terminal connected becomes aij > aij .
Repeat part (a).
(c) Suppose that when no terminals are connected to concentrator j there is
a given cost savings cj > 0. Can you still formulate the problem as a
minimum cost flow problem?

1.26

Consider a round-robin chess tournament involving n players that play each other
once. A win scores 1 for the winner and 0 for the loser, while a draw scores 1/2
Sec. 1.4 Notes, Sources, and Exercises 47

for each player. We are given a set of final scores (s1 , . . . , sn ) for the players, from
the range [0, n − 1], whose sum is n(n − 1)/2, and we want to check whether these
scores are feasible [for example, in a four-player tournament, a set of final scores
of (3, 3, 0, 0) is impossible]. Show that this is equivalent to checking feasibility of
some transportation problem.

1.27 (k-Color Problem)

Consider the k-color problem, which is to assign one out of k colors to each node
of a graph so that for every arc (i, j), nodes i and j have different colors.
(a) Suppose we want to choose the colors of countries in a world map so that
no two adjacent countries have the same color. Show that if the number of
available colors is k, the problem can be formulated as a k-color problem.
(b) Show that the k-color problem has a solution if and only if the number of
nodes can be partitioned in k or less disjoint subsets such that there is no
arc connecting a pair of nodes from the same subset.
(c) Show that when the graph is a tree, the 2-color problem has a solution.
Hint: First color some node i and then color the remaining nodes based on
their “distance” from i.
(d) Show that if each node has at most k − 1 neighbors, the k-color problem
has a solution.

1.28 (k-Coloring and Parallel Computation)

Consider the n-dimensional vector x = (x1 , . . . , xn ) and an iteration of the form

xj := fj (x), j = 1, . . . , n,

where f = (f1 , . . . , fn ) is a given function. The dependency graph of f has nodes

1, . . . , n and an arc set such that (i, j) is an arc if the function fj exhibits a
dependence on the component xi . Consider an ordering j1 , . . . , jn of the indices
1, . . . , n, and a partition of {j1 , . . . , jn } into disjoint subsets J1 , . . . , JM such that:
(1) For all k, if jk ∈ Jm , then jk+1 ∈ Jm ∪ · · · ∪ JM .
(2) If jp , jq ∈ Jm and p < q, then fjq does not depend on xjp .
Show that such an ordering and partition exist if and only if the nodes of the
dependency graph can be colored with M colors so that there exists no forward
cycle with all the nodes on the cycle having the same color. Note: This is
challenging (see Bertsekas and Tsitsiklis [1989], Section 1.2.4, for discussion and
analysis). An ordering and partition of this type can be used to execute Gauss-
Seidel iterations in M parallel steps.
48 Introduction Chap. 1

1.29 (Replacing Arc Costs with Reduced Costs)

Consider the minimum cost flow problem and let pj be a scalar price for each
node j. Show that if the arc cost coefficients aij are replaced by aij + pj − pi ,
we obtain a problem that is equivalent to the original (except for a scalar shift
in the cost function value).

1.30

Consider the assignment problem.

(a) Show that every k-person exchange can be accomplished with a sequence
of k − 1 successive two-person exchanges.
(b) In light of the result of part (a), how do you explain that a nonoptimal
assignment may not be improvable by any two-person exchange?

1.31 (Dual Cost Improvement Directions)

Consider the assignment problem. Let pj denote the price of object j, let T be a
subset of objects, and let

S = i | the maximum of aij − pj over j ∈ A(i)

is attained by some element of T .

Assume that:
(1) For each i ∈ S, the maximum of aij − pj over j ∈ A(i) is attained only by
elements of T .
(2) S has more elements than T .
Show that the direction d = (d1 , . . . , dn ), where dj = 1 if j ∈ T and dj = 0 if
j∈/ T , is a direction of dual cost improvement. Note: Directions of this type are
used by the most common dual cost improvement algorithms for the assignment
problem.

1.32

Use -CS to verify that the assignment of Fig. 1.18 is optimal and obtain a bound
on how far from optimal the given price vector is. State the dual problem and
verify the correctness of the bound by comparing the dual value of the price
vector with the optimal dual value.
Sec. 1.4 Notes, Sources, and Exercises 49

Value = C
1 11 p = C - 1/8
C
0 Figure 1.18: Graph of an assignment prob-
C lem. Objects 1 and 2 have value C for all
C p = C + 1/8 persons. Object 3 has value 0 for all per-
2 2
sons. Object prices are as shown. The
0 thick lines indicate the given assignment.
C C

3 0 3 p=0

1.33 (Generic Negative Cycle Algorithm)

Consider the following minimum cost flow problem


minimize aij xij
(i,j)∈A
 
subject to xij − xji = si , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

0 ≤ xij ≤ cij , ∀ (i, j) ∈ A,

and assume that the problem has at least one feasible solution. Consider first
the circulation case where si = 0 for all i ∈ N . Construct a sequence of flow
vectors x0 , x1 , . . . as follows: Start with x0 = 0. Given xk , stop if xk is optimal,
and otherwise find a simple cycle C k that is unblocked with respect to xk and
has negative cost (cf. Prop. 1.2). Increase (decrease) the flow of the forward
(backward, respectively) arcs of C k by the maximum possible increment.
(a) Show that the cost of xk+1 is smaller than the cost of xk by an amount
that is proportional to the cost of the cycle C k and to the increment of the
corresponding flow change.
(b) Assume that the flow increment at each iteration is greater or equal to
some scalar δ > 0. Show that the algorithm must terminate after a finite
number of iterations with an optimal flow vector. Note: The assumption
of existence of such a δ is essential (see Exercise 3.7 in Chapter 3).
(c) Extend parts (a) and (b) to the general case where we may have si = 0 for
some i, by converting the problem to the circulation format (a method for
doing this is given in Section 4.1.3).

1.34 (Integer Optimal Solutions of Min-Cost Flow Problems)

Consider the minimum cost flow problem of Exercise 1.33, where the upper
bounds cij are given positive integers and the supplies si are given integers.
Assume that the problem has at least one feasible solution. Show that there
exists an optimal flow vector that is integer. Hint: Show that the flow vectors
generated by the negative cycle algorithm of Exercise 1.33 are integer.
50 Introduction Chap. 1

1.35 (The Original Hamiltonian Cycle)

The origins of the traveling salesman problem can be traced (among others) to the
work of the Irish mathematician Sir William Hamilton. In 1856, he developed a
system of commutative algebra, which inspired a puzzle marketed as the “Icosian
Game.” The puzzle is to find a cycle that passes exactly once through each
of the 20 nodes of the graph shown in Fig. 1.19, which represents a regular
dodecahedron. Find a Hamiltonian cycle on this graph using as first four nodes
the ones marked 1-4 (all arcs are considered bidirectional).

1
2

4
3

Figure 1.19: Graph for the Icosian Game (cf. Exercise 1.35). The arcs and nodes
correspond to the edges and vertices of the regular dodecahedron, respectively.
The name “icosian” comes from the Greek word “icosi,” which means twenty.
Adjacent nodes of the dodecahedron correspond to adjacent faces of the regular
icosahedron.

1.36 (Hamiltonian Cycle on the Hypercube)

The hypercube of dimension n is a graph with 2n nodes, each corresponding to

an n-bit string where each bit is either a 0 or a 1. There is a bidirectional arc
connecting every pair of nodes whose n-bit strings differ by a single bit. Show
that for every n ≥ 2, the hypercube contains a Hamiltonian cycle. Hint: Use
induction.

1.37 (Hardy’s Theorem)

Let {a1 , . . . , an } and {b1 , . . . , bn } be monotonically nondecreasing sequences of

numbers. Consider the problem of associating with each i = 1, . . . , n a distinct
n
index ji in a way that maximizes a b . Formulate this as an assignment
i=1 i ji
problem and show that it is optimal to select ji = i for all i. Hint: Use the
complementary slackness conditions with prices defined by p1 = 0 and pk =
pk−1 + ak (bk − bk−1 ) for k = 2, . . . , n.
2

The Shortest Path Problem

Contents

2.1. Problem Formulation and Applications

2.2. A Generic Shortest Path Algorithm
2.3. Label Setting (Dijkstra) Methods
2.2.1. Performance of Label Setting Methods
2.3.2. The Binary Heap Method
2.3.3. Dial’s Algorithm
2.4. Label Correcting Methods
2.4.1. The Bellman-Ford Method
2.4.2. The D’Esopo-Pape Algorithm
2.4.3. The SLF and LLL Algorithms
2.4.4. The Threshold Algorithm
2.4.5. Comparison of Label Setting and Label Correcting
2.5. Single Origin/Single Destination Methods
2.5.1. Label Setting
2.5.2. Label Correcting
2.6. Auction Algorithms
2.7. Multiple Origin/Multiple Destination Methods
2.8. Notes, Sources, and Exercises

51
 52 The Shortest Path Problem Chap. 2

The shortest path problem is a classical and important combinatorial prob-

lem that arises in many contexts. We are given a directed graph (N , A)
with nodes numbered 1, . . . , N . Each arc (i, j) ∈ A has a cost or “length”
aij associated with it. The length of a forward path (i1 , i2 , . . . , ik ) is the
length of its arcs

k−1
ain in+1 .
n=1

This path is said to be shortest if it has minimum length over all forward
paths with the same origin and destination nodes. The length of a shortest
path is also called the shortest distance. The shortest path problem deals
with finding shortest distances between selected pairs of nodes. [Note that
here we are optimizing over forward paths; when we refer to a path (or a
cycle) in connection with the shortest path problem, we implicitly assume
that the path (or the cycle) is forward.]
The range of applications of the shortest path problem is very broad.
In the next section, we will provide some representative examples. We
will then develop a variety of algorithms. Most of these algorithms can be
viewed as primal cost or dual cost improvement algorithms for an appro-
priate special case of the minimum cost flow problem, as we will see later.
However, the shortest path problem is simple, so we will discuss it based
on first principles, and without much reference to cost improvement. This
serves a dual purpose. First, it provides an opportunity to illustrate some
basic graph concepts in the context of a problem that is simple and rich in
intuition. Second, it allows the early development of some ideas and results
that will be used later in a variety of other algorithmic contexts.

2.1 PROBLEM FORMULATION AND APPLICATIONS

The shortest path problem appears in a large variety of contexts. We

discuss a few representative applications.

Example 2.1. Routing in Data Networks

Data network communication involves the use of a network of computers

(nodes) and communication links (arcs) that transfer packets (groups of bits)
from their origins to their destinations. The most common method for se-
lecting the path of travel (or route) of packets is based on a shortest path
formulation. In particular, each communication link is assigned a positive
scalar which is viewed as its length. A shortest path routing algorithm routes
each packet along a minimum length (or shortest) path between the origin
and destination nodes of the packet.
There are several possibilities for selecting the link lengths. The sim-
plest is for each link to have unit length, in which case a shortest path is
Sec. 2.1 Problem Formulation and Applications 53

simply a path with minimum number of links. More generally, the length
of a link, may depend on its transmission capacity and its projected traffic
load. The idea here is that a shortest path should contain relatively few and
uncongested links, and therefore be desirable for routing. Sophisticated rout-
ing algorithms also allow the length of each link to change over time and to
depend on the prevailing congestion level of the link. Then a shortest path
may adapt to temporary overloads and route packets around points of con-
gestion. Within this context, the shortest path routing algorithm operates
continuously, solving the shortest path problem with lengths that vary over
time.
A peculiar feature of shortest path routing algorithms is that they are
often implemented using distributed and asynchronous communication and
computation. In particular, each node of the communication network mon-
itors the traffic conditions of its adjacent links, calculates estimates of its
shortest distances to various destinations, and passes these estimates to other
nodes who adjust their own estimates, etc. This process is based on stan-
dard shortest path algorithms that will be discussed in this chapter, but it
is also executed asynchronously, and with out-of-date information because of
communication delays between the nodes. Despite this fact, it turns out that
these distributed asynchronous algorithms maintain much of the validity of
their synchronous counterparts (see the textbooks by Bertsekas and Tsitsiklis
[1989], and Bertsekas and Gallager [1992] for related analysis).

There is an important connection between shortest path problems

and problems of deterministic discrete-state dynamic programming, which
involve sequential decision making over a finite number of time periods.
The following example shows that dynamic programming problems can be
formulated as shortest path problems. The reverse is also possible; that is,
any shortest path problem can be formulated as a dynamic programming
problem (see e.g., Bertsekas [1995a], Ch. 2).

Example 2.2. Dynamic Programming

Here we have a discrete-time dynamic system involving N stages. The state

of the system at the start of the kth stage is denoted by xk and takes values
in a given finite set, which may depend on the index k. The initial state x0 is
given. During the kth stage, the state of the system changes from xk to xk+1
according to an equation of the form

xk+1 = fk (xk , uk ), (2.1)

where uk is a control that takes values from a given finite set, which may
depend on the index k. This transition involves a cost gk (xk , uk ). The final
transition from xN −1 to xN , involves an additional terminal cost G(xN ).
Here, the functions fk , gk , and G are given.
Given a control sequence (u0 , . . . , uN −1 ), the corresponding state se-
quence (x0 , . . . , xN ) is determined from the given initial state x0 and the
system of Eq. (2.1). The objective in dynamic programming is to find a
54 The Shortest Path Problem Chap. 2

control sequence and a corresponding state sequence such that the total cost


N −1

G(xN ) + gk (xk , uk )
k=0

is minimized.
For an example, consider an inventory system that operates over N
time periods, and let xk and uk denote the number of items held in stock and
number of items purchased at the beginning of period k, respectively. We
require that uk be an integer from a given range [0, rk ]. We assume that the
stock evolves according to the equation

xk+1 = xk + uk − vk ,

where vk is a known integer demand for period k; this is the system equa-
tion [cf. Eq. (2.1)]. A negative xk here indicates unsatisfied demand that is
backordered. A common type of cost used in inventory problems has the form

gk (xk , uk ) = hk (xk ) + ck uk ,

where ck is a given cost per unit stock at period k, and hk (xk ) is a cost either
for carrying excess inventory (xk > 0) or for backordering demand (xk < 0).
For example hk (xk ) = max{ak xk , −bk xk } or hk (xk ) = dk x2k , where ak , bk , and
dk are positive scalars, are both reasonable choices for cost function. Finally,
we could take G(xN ) = 0 to indicate that the final stock xN has no value
[otherwise G(xN ) indicates the cost (or negative salvage value) of xN ]. The
objective in this problem is roughly to determine the sequence of purchases
over time to minimize the costs of excess inventory and backordering demand
over the N time periods.
To convert the dynamic programming problem to a shortest path prob-
lem, we introduce a graph such as the one of Fig. 2.1, where the arcs corre-
spond to transitions between states at successive stages and each arc has a
cost associated with it. To handle the final stage, we also add an artificial
terminal node t. Each state xN at stage N is connected to the terminal node
t with an arc having cost G(xN ). Control sequences correspond to paths
originating at the initial state x0 and terminating at one of the nodes corre-
sponding to the final stage N . The optimal control sequence corresponds to a
shortest path from node x0 to node t. For an extensive treatment of dynamic
programming and associated shortest path algorithms we refer to Bertsekas
[1995a].

Shortest path problems arise often in contexts of scheduling and se-

quencing. The following two examples are typical.

Example 2.3. Project Management

Consider the planning of a project involving several activities, some of which

must be completed before others can begin. The duration of each activity is
Sec. 2.1 Problem Formulation and Applications 55

x2 xN- 1

Terminal Arcs
x1 . . . xN with Cost Equal
to Terminal Cost

. . .
Initial State
x0 t
Artificial Terminal
. . . Node

. . .

Stage 0 Stage 1 Stage 2 . . . Stage N - 1 Stage N

Figure 2.1: Converting a deterministic finite-state N -stage dynamic program-

ming problem to a shortest path problem. Nodes correspond to states. An arc
with start and end nodes xk and xk+1 , respectively, corresponds to a transition
of the form xk+1 = fk (xk , uk ). The length of this arc is equal to the cost of
the corresponding transition gk (xk , uk ). The problem is equivalent to finding a
shortest path from the initial state/node x0 to the artificial terminal node t. Note
that the state space and the possible transitions between states may depend on
the stage index k.

known in advance. We want to find the time required to complete the project,
as well as the critical activities, those that even if slightly delayed will result
in a corresponding delay of completion of the overall project.
The problem can be represented by a graph where nodes represent
completion of some phase of the project (cf. Fig. 2.2). An arc (i, j) represents
an activity that starts once phase i is completed and has known duration
tij > 0. A phase (node) j is completed when all activities or arcs (i, j) that
are incoming to j are completed. Two special nodes 1 and N represent the
start and end of the project, respectively. Node 1 has no incoming arcs,
while node N has no outgoing arcs. Furthermore, there is at least one path
from node 1 to every other node. An important characteristic of an activity
network is that it is acyclic. This is inherent in the problem formulation and
the interpretation of nodes
as phase completions. 
For any path p = (1, j1 ), (j1 , j2 , ), . . . , (jk , i) from node 1 to a node
i, let Dp be the duration of the path defined as the sum of durations of its
activities; that is,

Dp = tij1 + tj1 j2 + · · · + tjk i .

Then the time Ti required to complete phase i is

Ti = max Dp .
paths p
from 1 to i

The maximum above is attained by some path because there can be only a
finite number of paths from 1 to i, since the network is acyclic. Thus to find
56 The Shortest Path Problem Chap. 2

Order 3 Transport
Material Material
Start 2 End
2 Construction
Train 4 5
1 1
Personnel 4
3 2
Hire Train
Personnel Personnel
2

Figure 2.2: Example graph of an activity network. Arcs (i, j) represent

activities and are labeled by the corresponding duration tij . Nodes represent
completion of some phase of the project. A phase is completed if all activities
associated with incoming arcs at the corresponding node are completed. The
project is completed when all phases are completed. The project duration
time is the longest sum of arc durations over paths that start at node 1 and
end at node 5. The path of longest duration, also called a critical path, is
shown with thick line. Because the graph is acyclic, finding this path is a
shortest path problem with the length of each arc (i, j) being −tij . Activities
on the critical path have the property that if any one of them is delayed, a
corresponding delay of completion of the overall project will result.

Ti , we should find the longest path from 1 to i. Because the graph is acyclic,
this problem may also be viewed as a shortest path problem with the length
of each arc (i, j) being −tij . In particular, finding the duration of the project
is equivalent to finding the shortest path from 1 to N . For further discussion
of project management problems, we refer to the literature, e.g., the textbook
by Elmaghraby [1978].

Example 2.4. The Paragraphing Problem

This problem arises in a word processing context, where we want to break

down a given paragraph consisting of N words into lines for “optimal” ap-
pearance and readability. Suppose that we have a heuristic rule, which assigns
to any sequence of words a cost that expresses the undesirability of grouping
these words together in a line. Based on such a rule, we can assign a cost
cij to a line starting with word i and ending with word j − 1 of the given
paragraph. An optimally divided paragraph is one for which the sum of the
costs of its lines is minimal.
We can formulate this as a shortest path problem. There are N nodes,
which correspond to the N words of the paragraph, and there is an arc (i, j)
with cost cij connecting any two words i and j with i < j. The arcs of the
shortest path from node/word 1 to node/word N correspond to the lines of
the optimally broken down paragraph.
 Sec. 2.2 A Generic Shortest Path Algorithm 57

The exercises contain a number of additional examples that illustrate

the broad range of applications of the shortest path problem.

2.2 A GENERIC SHORTEST PATH ALGORITHM

The shortest path problem can be posed in a number of ways; for example,
finding a shortest path from a single origin to a single destination, or finding
a shortest path from each of several origins to each of several destinations.
We focus initially on problems with a single origin and many destinations.
For concreteness, we take the origin node to be node 1. The arc lengths aij
are given scalars. They may be negative and/or noninteger, although on
occasion we will assume in our analysis that they are nonnegative and/or
integer, in which case we will state so explicitly.
In this section, we develop a broad class of shortest path algorithms
for the single origin/all destinations problem. These algorithms maintain
and adjust a vector (d1 , d2 , . . . , dN ), where each dj , called the label of node
j, is either a scalar or ∞. The use of labels is motivated by a simple
optimality condition, which is given in the following proposition.

Proposition 2.1: Let d1 , d2 , . . . , dN be scalars satisfying

dj ≤ di + aij , ∀ (i, j) ∈ A, (2.2)

and let P be a path starting at a node i1 and ending at a node ik . If

dj = di + aij , for all arcs (i, j) of P , (2.3)

then P is a shortest path from i1 to ik .

Proof: By adding Eq. (2.3) over the arcs of P , we see that the length of
P is equal to the difference dik − di1 of labels of the end node and start
node of P . By adding Eq. (2.2) over the arcs of any other path P
starting
at i1 and ending at ik , we see that the length of P
must be no less than
dik − di1 . Therefore, P is a shortest path. Q.E.D.

The conditions (2.2) and (2.3) are called the complementary slackness
(CS) conditions for the shortest path problem. This terminology is moti-
vated by the connection of the shortest path problem with the minimum
cost flow problem (cf. Section 1.2.1); we will see in Chapter 4 that the CS
conditions of Prop. 2.1 are a special case of a general optimality condition
(also called CS condition) for the equivalent minimum cost flow problem
58 The Shortest Path Problem Chap. 2

(in fact they are a special case of a corresponding CS condition for general
linear programs; see e.g., Bertsimas and Tsitsiklis [1997], Dantzig [1963]).
Furthermore, we will see that the scalars di in Prop. 2.1 are related to dual
variables.
Let us now describe a prototype shortest path method that contains
several interesting algorithms as special cases. In this method, we start
with some vector of labels (d1 , d2 , . . . , dN ), we successively select arcs (i, j)
that violate the CS condition (2.2), i.e., dj > di + aij , and we set

dj := di + aij .

This is continued until the CS condition dj ≤ di + aij is satisfied for all

arcs (i, j).
A key idea is that, in the course of the algorithm, di can be interpreted
for all i as the length of some path Pi from 1 to i.† Therefore, if dj > di +aij
for some arc (i, j), the path obtained by extending path Pi by arc (i, j),
which has length di + aij , is a better path than the current path Pj , which
has length dj . Thus, the algorithm finds successively better paths from the
origin to various destinations.
Instead of selecting arcs in arbitrary order to check violation of the CS
condition, it is usually most convenient and efficient to select nodes, one-at-
a-time according to some order, and simultaneously check violation of the
CS condition for all of their outgoing arcs. The corresponding algorithm,
referred to as generic, maintains a list of nodes V , called the candidate
list, and a vector of labels (d1 , d2 , . . . , dN ), where each dj is either a real
number or ∞. Initially,
V = {1},
d1 = 0, di = ∞, ∀ i = 1.
The algorithm proceeds in iterations and terminates when V is empty. The
typical iteration (assuming V is nonempty) is as follows:

Iteration of the Generic Shortest Path Algorithm

Remove a node i from the candidate list V . For each outgoing arc
(i, j) ∈ A, if dj > di + aij , set

dj := di + aij

and add j to V if it does not already belong to V .

† In the case of the origin node 1, we will interpret the label d1 as either the
length of a cycle that starts and ends at 1, or (in the case d1 = 0) the length of
the trivial “path” from 1 to itself.
Sec. 2.2 A Generic Shortest Path Algorithm 59

2 2
3

Origin 1 1 1 4

1 3
3

Iteration # Candidate List V Node Labels Node out of V

1 {1} (0, ∞, ∞, ∞) 1
2 {2, 3} (0, 3, 1, ∞) 2
3 {3, 4} (0, 3, 1, 5) 3
4 {4, 2} (0, 2, 1, 4) 4
5 {2} (0, 2, 1, 4) 2
Ø (0, 2, 1, 4)

Figure 2.3: Illustration of the generic shortest path algorithm. The numbers
next to the arcs are the arc lengths. Note that node 2 enters the candidate list
twice. If in iteration 2 node 3 was removed from V instead of node 2, each node
would enter V only once. Thus, the order in which nodes are removed from V is
significant.

It can be seen that, in the course of the algorithm, the labels are
monotonically nonincreasing. Furthermore, we have
di < ∞ ⇐⇒ i has entered V at least once.
Figure 2.3 illustrates the algorithm. The following proposition gives its
main properties.

Proposition 2.2: Consider the generic shortest path algorithm.

(a) At the end of each iteration, the following conditions hold:
(i) If dj < ∞, then dj is the length of some path that starts
at 1 and ends at j.
(ii) If i ∈
/ V , then either di = ∞ or else

dj ≤ di + aij , ∀ j such that (i, j) ∈ A.

60 The Shortest Path Problem Chap. 2

(b) If the algorithm terminates, then upon termination, for all j with
dj < ∞, dj is the shortest distance from 1 to j and

min(i,j)∈A {di + aij } if j = 1,
dj = (2.4)
0 if j = 1.

Furthermore, upon termination we have dj = ∞ if and only if

there is no path from 1 to j.
(c) If the algorithm does not terminate, then there exists some node
j and a sequence of paths that start at 1, end at j, and have
lengths that diverge to −∞.
(d) The algorithm terminates if and only if there is no path that
starts at 1 and contains a cycle with negative length.

Proof: (a) We prove (i) by induction on the iteration count. Indeed, (i)
holds at the end of the first iteration since the nodes j = 1 with dj < ∞
are those for which (1, j) is an arc and their labels are dj = a1j , while for
the origin 1, we have d1 = 0, which by convention is viewed as the length
of the trivial “path” from 1 to itself. Suppose that (i) holds at the start
of some iteration at which the node removed from V is i. Then di < ∞
(which is true for all nodes of V by the rules of the algorithm), and (by the
induction hypothesis) di is the length of some path Pi starting at 1 and
ending at i. When a label dj changes as a result of the iteration, dj is set
to di + aij , which is the length of the path consisting of Pi followed by arc
(i, j). Thus property (i) holds at the end of the iteration, completing the
induction proof.
To prove (ii), note that for any i, each time i is removed from V , the
condition dj ≤ di + aij is satisfied for all (i, j) ∈ A by the rules of the
algorithm. Up to the next entrance of i into V , di stays constant, while
the labels dj for all j with (i, j) ∈ A cannot increase, thereby preserving
the condition dj ≤ di + aij .
(b) We first introduce the sets

I = {i | di < ∞ upon termination},

I = {i | di = ∞ upon termination},

and we show that we have j ∈ I if and only if there is no path from 1 to j.

Indeed, if i ∈ I, we have di < ∞ and therefore dj < ∞ for all j such that
(i, j) is an arc in view of condition (ii) of part (a), so that j ∈ I. It follows
that there is no path from any node of I (and in particular, node 1) to
any node of I. Conversely, if there is no path from 1 to j, it follows from
Sec. 2.2 A Generic Shortest Path Algorithm 61

condition (i) of part (a) that we cannot have dj < ∞ upon termination, so
j ∈ I.
We show now that for all j ∈ I, upon termination, dj is the shortest
distance from 1 to j and Eq. (2.4) holds. Indeed, conditions (i) and (ii) of
part (a) imply that upon termination we have, for all i ∈ I,

dj ≤ di + aij , ∀ j such that (i, j) ∈ A, (2.5)

while di is the length of some path from 1 to i, denoted Pi . Fix a node

m ∈ I, and consider any path P from 1 to m. By adding the condition
(2.5) over the arcs of P , we see that the length of P is no less than dm − d1 ,
which is less or equal to dm (we have d1 ≤ 0, since initially d1 = 0 and all
node labels are monotonically nonincreasing). Hence Pm is a shortest path
from 1 to m and the shortest distance is dm . Furthermore, the equality
dj = di + aij must hold for all arcs (i, j) on the shortest paths Pm , m ∈ I,
implying that dj = min(i,j)∈A {di + aij } for all j ∈ I with j = 1, while
d1 = 0.
(c) If the algorithm never terminates, some label dj must decrease strictly
an infinite number of times, generating a corresponding sequence of distinct
paths Pj as per condition (i) of part (a). Each of these paths can be
decomposed into a simple path from 1 to j plus a collection of simple
cycles, as in Exercise 1.4 of Chapter 1. Since the number of simple paths
from 1 to j is finite, and the length of Pj is monotonically decreasing, it
follows that Pj eventually must involve a cycle with negative length. By
replicating this cycle a sufficiently large number of times, one can obtain
paths from 1 to j with arbitrarily small length.
(d) Using part (c), we have that the algorithm will terminate if and only if
there is a lower bound on the length of all paths that start at node 1. Thus,
the algorithm will terminate if and only if there is no path that starts at
node 1 and contains a cycle with negative length. Q.E.D.

When some arc lengths are negative, Prop. 2.2 points to a way to
detect existence of a path that starts at the origin 1 and contains a cycle
of negative length. If such a path exists, it can be shown under mild
assumptions that the label of at least one node will diverge to −∞ (see
Exercise 2.32). We can thus monitor whether for some j we have

dj < (N − 1) min aij .

(i,j)∈A

When this condition occurs, the path from 1 to j whose length is equal to
dj [as per Prop. 2.2(a)] must contain a negative cycle [if it were simple, it
would consist of at most N − 1 arcs, and its length could not be smaller
than (N − 1) min(i,j)∈A aij ; a similar argument would apply if it were not
simple but it contained only cycles of nonnegative length].
62 The Shortest Path Problem Chap. 2

Bellman’s Equation and Shortest Path Construction

When all cycles have nonnegative length and there exists a path from node
1 to every node j, then Prop. 2.2 shows that the generic algorithm termi-
nates and that, upon termination, all labels are equal to the corresponding
shortest distances, and satisfy d1 = 0 and

dj = min {di + aij }, ∀ j = 1. (2.6)

(i,j)∈A

This is known as Bellman’s equation and it has an intuitive meaning: it

indicates that the shortest distance from 1 to j is obtained by optimally
choosing the predecessor i of node j in order to minimize the sum of the
shortest distance from 1 to i and the length of arc (i, j). It also indicates
that if Pj is a shortest path from 1 to j, and a node i belongs to Pj , then
the portion of Pj from 1 to i, is a shortest path from 1 to i.
From Bellman’s equation, we can obtain the shortest paths (in addi-
tion to the shortest path lengths) if all cycles not including node 1 have
strictly positive length. To do this, select for each j = 1 one arc (i, j)
that attains the minimum in dj = min(i,j)∈A {di + aij } and consider the
subgraph consisting of these N − 1 arcs; see Fig. 2.4. To find the short-
est path to any node j, start from j and follow the corresponding arcs of
the subgraph backward until node 1 is reached. Note that the same node
cannot be reached twice before node 1 is reached, since a cycle would be
formed that, on the basis of Eqs. (2.6), would have zero length. [To see
this, let (i1 , i2 , . . . , ik , i1 ) be the cycle and add the equations

di1 = di2 + ai2 i1

...

dik−1 = dik + aik ik−1

dik = di1 + ai1 ik

obtaining ai2 i1 + · · · + aik ik−1 + ai1 ik = 0.] Since the subgraph is connected
and has N − 1 arcs, it must be a spanning tree. We call this subgraph a
shortest path spanning tree, and we note its special structure: it has a root
(node 1) and every arc of the tree is directed away from the root. The
preceding argument can also be used to show that Bellman’s equation has
no solution other than the shortest distances; see Exercise 2.5.
A shortest path spanning tree can also be constructed in the process
of executing the generic shortest path algorithm by recording the arc (i, j)
every time dj is decreased to di + aij ; see Exercise 2.4.
Sec. 2.2 A Generic Shortest Path Algorithm 63

Figure 2.4: Example of construction

of shortest path spanning tree. The arc
d 2= 2 lengths are shown next to the arcs, and
2 the shortest distances are shown next
4 1
Origin to the nodes. For each j
= 1, we select
an arc (i, j) such that
1 1 1 4 d =3
4
dj = di + aij
1 3
3
d 3= 1 and we form the shortest path spanning
tree. The arcs selected in this example
are (1, 3), (3, 2), and (2, 4).

Advanced Initialization

The generic algorithm need not be started with the initial conditions

V = {1}, d1 = 0, di = ∞, ∀ i = 1,

in order to work correctly. Any set of labels (d1 , . . . , dN ) and candidate

list V can be used initially, as long as they satisfy the conditions of Prop.
2.2(a). It can be seen that the proof of the remaining parts of Prop. 2.2 go
through under these conditions.
In particular, the algorithm works correctly if the labels and the can-
didate list are initialized so that d1 = 0 and:
(a) For each node i, di is either ∞ or else it is the length of a path from
1 to i.
(b) The candidate list V contains all nodes i such that

di + aij < dj for some (i, j) ∈ A. (2.7)

This kind of initialization is very useful in reoptimization contexts,

where we have to solve a large number of similar problems that differ
slightly from each other; for example they may differ by just a few arc
lengths or they may have a slightly different node set. The lengths of the
shortest paths of one problem can be used as the starting labels for another
problem, and substantial computational savings may be obtained, because
it is likely that many of the nodes will maintain their shortest path lengths
and will never enter V .
Another important situation where an advanced initialization is very
useful arises if, by using heuristics or an available solution of a similar
shortest path problem, we can construct a set of “good” paths from node 1
to the other nodes. Then we can use the lengths of these paths as the initial
labels in the generic shortest path algorithm and start with a candidate list
consisting of the nodes where the CS condition is violated [cf. Eq. (2.7)].
64 The Shortest Path Problem Chap. 2

Finally, let us note another technique that is sometimes useful in

reoptimization settings. Suppose that we have some scalars δ1 , . . . , δN and
we change the arc lengths to
âij = aij + δi − δj .
Then it can be seen that the length of any path from a node m to a node
n will be increased by δm − δn , while the shortest paths will be unaffected.
Thus it may be advantageous to use the modified arc lengths âij instead
of the original lengths aij , if this will enhance the application of a suitable
shortest path algorithm. For example, we may be able with proper choice
of δi , to reduce the arc cost range max(i,j) |âij | (this is helpful in some
algorithms) or to make âij nonnegative (see Section 2.7 for an application
of this idea).

Implementations of the Generic Algorithm

There are many implementations of the generic algorithm. They differ in

how they select the node to be removed from the candidate list V , and
they are broadly divided into two categories:
(a) Label setting methods. In these methods, the node i removed from
V is a node with minimum label. Under the assumption that all arc
lengths are nonnegative, these methods have a remarkable property:
each node will enter V at most once, as we will show shortly; its label
has its permanent or final value at the first time it is removed from
V . The most time-consuming part of these methods is calculating
the minimum label node in V at each iteration; there are several
implementations, that use a variety of creative procedures to obtain
this minimum.
(b) Label correcting methods. In these methods the choice of the node i
removed from V is less sophisticated than in label setting methods,
and requires less calculation. However, a node may enter V multiple
times.
There are several worst-case complexity bounds for label setting and
label correcting methods. The best bounds for the case of nonnegative arc
lengths correspond to label setting methods. The best practical methods,
however, are not necessarily the ones with the best complexity bounds, as
will be discussed in the next two sections.
In practice, when the arc lengths are nonnegative, the best label set-
ting methods and the best label correcting methods are competitive. As a
general rule, a sparse graph favors the use of a label correcting over a label
setting method for reasons that will be explained later (see the discussion at
the end of Section 2.4). An important advantage of label correcting meth-
ods is that they are more general, since they do not require nonnegativity
of the arc lengths.
 Sec. 2.3 Label Setting (Dijkstra) Methods 65

2.3 LABEL SETTING (DIJKSTRA) METHODS

In this section we discuss various implementations of the label setting ap-

proach. The prototype label setting method, first published by Dijkstra
[1959] but also discovered independently by several other researchers, is
the special case of the generic algorithm where the node i removed from
the candidate list V at each iteration has minimum label, that is,
di = min dj .
j∈V

For convenient reference, let us state this method explicitly.

Initially, we have
V = {1},
d1 = 0, di = ∞, ∀ i = 1.
The method proceeds in iterations and terminates when V is empty. The
typical iteration (assuming V is nonempty) is as follows:

Iteration of the Label Setting Method

Remove from the candidate list V a node i such that

di = min dj .
j∈V

For each outgoing arc (i, j) ∈ A, if dj > di + aij , set

dj := di + aij

and add j to V if it does not already belong to V .

Figure 2.5 illustrates the label setting method. Some insight into the
method can be gained by considering the set W of nodes that have already
been in V but are not currently in V :
W = {i | di < ∞, i ∈
/ V }.
We will prove later, in Prop. 2.3(a), that as a consequence of the policy of
removing from V a minimum label node, W contains nodes with “small”
labels throughout the algorithm, in the sense that
dj ≤ d i , if j ∈ W and i ∈
/ W. (2.8)
Using this property and the assumption aij ≥ 0, it can be seen that when
a node i is removed from V , we have, for all j ∈ W for which (i, j) is an
arc,
dj ≤ di + aij .
66 The Shortest Path Problem Chap. 2

2 0
2 1
Origin 1 1 1 4 5
0
1 3

Iteration # Candidate List V Node Labels Node out of V

1 {1} (0, ∞, ∞, ∞, ∞) 1
2 {2, 3} (0, 2, 1, ∞, ∞) 3
3 {2, 4} (0, 2, 1, 4, ∞) 2
4 {4, 5} (0, 2, 1, 3, 2) 5
5 {4} (0, 2, 1, 3, 2) 4
Ø (0, 2, 1, 3, 2)

Figure 2.5: Example illustrating the label setting method. At each iteration,
the node with the minimum label is removed from V . Each node enters V only
once.

Hence, once a node enters W , it stays in W and its label does not change
further. Thus, W can be viewed as the set of permanently labeled nodes,
that is, the nodes that have acquired a final label, which by Prop. 2.2, must
be equal to their shortest distance from the origin.
The following proposition makes the preceding argument precise and
proves some additional facts.

Proposition 2.3: Assume that all arc lengths are nonnegative.

(a) For any iteration of the label setting method, the following hold
for the set
W = {i | di < ∞, i ∈/ V }. (2.9)

(i) No node belonging to W at the start of the iteration will

enter the candidate list V during the iteration.
(ii) At the end of the iteration, we have di ≤ dj for all i ∈ W
and j ∈
/ W.
Sec. 2.3 Label Setting (Dijkstra) Methods 67

(iii) For each node j, consider simple paths that start at 1, end
at j, and have all their other nodes in W at the end of the
iteration. Then the label dj at the end of the iteration is
equal to the length of the shortest of these paths (dj = ∞
if no such path exists).
(b) The label setting method will terminate, and all nodes with a
final label that is finite will be removed from the candidate list
V exactly once in order of increasing shortest distance from node
1; that is, if the final labels of i and j are finite and satisfy di < dj ,
then i will be removed before j.

Proof: (a) Properties (i) and (ii) will be proved simultaneously by induc-
tion on the iteration count. Clearly (i) and (ii) hold for the initial iteration
at which node 1 exits V and enters W .
Suppose that (i) and (ii) hold for iteration k − 1, and suppose that
during iteration k, node i satisfies di = minj∈V dj and exits V . Let W
and W be the set of Eq. (2.9) at the start and at the end of iteration k,
respectively. Let dj and dj be the label of each node j at the start and at
the end of iteration k, respectively. Since by the induction hypothesis we
have dj ≤ di for all j ∈ W , and aij ≥ 0 for all arcs (i, j), it follows that
dj ≤ di + aij for all arcs (i, j) with j ∈ W . Hence, a node j ∈ W cannot
enter V at iteration k. This completes the induction proof of property (i),
and shows that
W = W ∪ {i}.

Thus, at iteration k, the only labels that may change are the labels dj
of nodes j ∈/ W such that (i, j) is an arc; the label dj at the end of the
iteration will be min{dj , di + aij }. Since aij ≥ 0, di ≤ dj for all j ∈
/ W,
and di = di , we must have di ≤ dj for all j ∈ / W . Since by the induction
hypothesis we have dm ≤ di and dm = dm for all m ∈ W , it follows that
dm ≤ dj for all m ∈ W and j ∈ / W . This completes the induction proof of
property (ii).
To prove property (iii), choose any node j and consider the subgraph
consisting of the nodes W ∪ {j} together with the arcs that have both
end nodes in W ∪ {j}. Consider also a modified shortest path problem
involving this subgraph, and the same origin and arc lengths as in the
original shortest path problem. In view of properties (i) and (ii), the label
setting method applied to the modified shortest path problem yields the
same sequence of nodes exiting V and the same sequence of labels as when
applied to the original problem up to the current iteration. By Prop.
2.2, the label setting method for the modified problem terminates with the
labels equal to the shortest distances of the modified problem at the current
68 The Shortest Path Problem Chap. 2

iteration. This means that the labels at the end of the iteration have the
property stated in the proposition.
(b) Since there is no cycle with negative length, by Prop. 2.2(d), we see
that the label setting method will terminate. At each iteration the node
removed from V is added to W , and according to property (i) (proved
above), no node from W is ever returned to V . Therefore, each node
with a final label that is finite will be removed from V and simultaneously
entered in W exactly once, and, by the rules of the algorithm, its label
cannot change after its entrance in W . Property (ii) then shows that each
new node added to W has a label at least as large as the labels of the nodes
already in W . Therefore, the nodes are removed from V in the order stated
in the proposition. Q.E.D.

2.3.1 Performance of Label Setting Methods

In label setting methods, the candidate list V is typically maintained with

the help of some data structure that facilitates the removal and the addition
of nodes, and also facilitates finding the minimum label node from the list.
The choice of data structure is crucial for good practical performance as
well as for good theoretical worst-case performance.
To gain some insight into this, we first consider a somewhat naive
implementation that will serve as a yardstick for comparison. By Prop.
2.3, there will be exactly N iterations, and in each of these, the candidate
list V will be searched for a minimum label node. Suppose this is done
by examining all nodes in sequence, checking whether they belong to V ,
and finding one with minimum label among those who do. Searching V
in this way requires O(N ) operations per iteration, for a total of O(N 2 )
operations. Also during the algorithm, we must examine each arc (i, j)
exactly once to check whether the condition dj > di + aij holds, and to set
dj := di + aij if it does. This requires O(A) operations, which is dominated
by the preceding O(N 2 ) estimate.
The O(A) operation count for arc examination is unavoidable and
cannot be reduced [each arc (i, j) must be checked at least once just to ver-
ify the optimality condition dj ≤ di + aij ]. However, the O(N 2 ) operation
count for minimum label searching can be reduced considerably by using
appropriate data structures. The best estimates of the worst-case√running
time that have been thus obtained are O(A+N log N ) and O(A+N log C),
where C is the arc length range C = max(i,j)∈A aij ; see Fredman and Tar-
jan [1984], and Ahuja, Mehlhorn, Orlin, and Tarjan [1990]. On the basis
of present experience, however, the implementations that perform best in
practice have considerable less favorable running time estimates. The ex-
planation for this is that the O(·) estimates involve a different constant for
each method and also correspond to worst-case problem instances. Thus,
the worst-case complexity estimates may not provide a reliable practical
Sec. 2.3 Label Setting (Dijkstra) Methods 69

comparison of various methods. We now discuss two of the most popular

implementations of the label setting method.

2.3.2 The Binary Heap Method

Here the nodes are organized as a binary heap on the basis of label values
and membership in V ; see Fig. 2.6. The node at the top of the heap is the
node of V that has minimum label, and the label of every node in V is no
larger than the labels of all the nodes that are in V and are its descendants
in the heap. Nodes that are not in V may be in the heap but may have no
descendants that are in V .

Label = 1

Label = 4 Label = 2

Label = 5 Node not in V Label = 2

Label = 4

Node not in V Label = 7 Node not in V Label = 4 Label = 6 Node not in V

Figure 2.6: A binary heap organized on the basis of node labels is a binary
balanced tree such that the label of each node of V is no larger than the labels of
all its descendants that are in V . Nodes that are not in V may have no descendants
that are in V . The topmost node, called the root, has the minimum label. The
tree is balanced in that the numbers of arcs in the paths from the root to any
nodes with no descendants differ by at most 1. If the label of some node decreases,
the node must be moved upward toward the root, requiring O(log N ) operations.
[It takes O(1) operations to compare the label of a node i with the label of one
of its descendants j, and to interchange the positions of i and j if the label of j
is smaller. Since there are log N levels in the tree, it takes at most log N such
comparisons and interchanges to move a node upward to the appropriate position
once its label is decreased.] Similarly, when the topmost node is removed from V ,
moving the node downward to the appropriate level in the heap requires at most
log N steps and O(log N ) operations. (Each step requires the interchange of the
position of the node and the position of one of its descendants. The descendant
must be in V for the step to be executed; if both descendants are in V , the one
with smaller label is selected.)

At each iteration, the top node of the heap is removed from V . Fur-
thermore, the labels of some nodes already in V may decrease, so these
may have to be repositioned in the heap; also, some other nodes may enter
70 The Shortest Path Problem Chap. 2

V for the first time and have to be inserted in the heap at the right place.
It can be seen that each of these removals, repositionings, and insertions
can be done in O(log N ) time. There are a total of N removals and N
node
 insertions, so  the number of operations for maintaining the heap is
O (N + R) log N , where R is the total number of node repositionings.
There is at most one repositioning per arc, since each arc is examined at
most once, so we have R ≤ A and the total operation count for maintaining
the heap is O(A log N ). This dominates the O(A) operation count to ex-
amine all arcs, so the worst-case running time of the method is O(A log N ).
On the other hand, practical experience indicates that the number of node
repositionings R is usually a small multiple of N , and considerably less
than the upper bound A. Thus, the running time of the method in prac-
tice typically grows approximately like O(A + N log N ).

2.3.3 Dial’s Algorithm

This algorithm, due to Dial [1969], requires that all arc lengths are non-
negative integers. It uses a naive yet often surprisingly effective method
for finding the minimum label node in V . The idea is to maintain for every
possible label value, a list of the nodes that have that value. Since every
finite label is equal to the length of some path with no cycles [Prop. 2.3(a),
part (iii)], the possible label values range from 0 to (N − 1)C, where

C = max aij .
(i,j)∈A

Thus, we may scan the (N − 1)C + 1 possible label values (in ascending
order) and look for a label value with nonempty list, instead of scanning
the candidate list V .
To visualize the algorithm, it is useful to think of each integer in
the range [0, (N − 1)C] as some kind of container, referred to as a bucket.
Each bucket b holds the nodes with label equal to b. Tracing steps, we see
that the method starts with the origin node 1 in bucket 0 and all other
buckets empty. At the first iteration, each node j with (1, j) ∈ A enters
the candidate list V and is inserted in bucket a1j . After we are done with
bucket 0, we proceed to check bucket 1. If it is nonempty, we repeat the
process, removing from V all nodes with label 1 and moving other nodes
to smaller numbered buckets as required; if not, we check bucket 2, and so
on. Figure 2.7 illustrates the method with an example.
Let us now consider the efficient implementation of the algorithm. We
first note that a doubly linked list (see Fig. 2.8) can be used to maintain the
set of nodes belonging to a given bucket, so that checking the emptiness of
a bucket and inserting or removing a node from a bucket are easy, requiring
O(1) operations. With such a data structure, the time required for mini-
mum label node searching is O(N C), and the time required for adjusting
node labels and repositioning nodes between buckets is O(A). Thus the
Sec. 2.3 Label Setting (Dijkstra) Methods 71

2
2 0

Origin 1 1 1 1 5
4 0
1
3 3

Iter. Cand. Node Buck. Buck. Buck. Buck. Buck. Out

# List V Labels 0 1 2 3 4 of V

1 {1} (0, ∞, ∞, ∞, ∞) 1 – – – – 1
2 {2, 3} (0, 2, 1, ∞, ∞) 1 3 2 – – 3
3 {2, 4} (0, 2, 1, 4, ∞) 1 3 2 – 4 2
4 {4, 5} (0, 2, 1, 3, 2) 1 3 2,5 4 – 5
5 {4} (0, 2, 1, 2, 2) 1 3 2,4,5 – – 4
Ø (0, 2, 1, 2, 2) 1 3 2,4,5 – –

Figure 2.7: An example illustrating Dial’s method.

overall running time is O(A + N C). The algorithm is pseudopolynomial,

but for small values of C (much smaller than N ) it performs very well in
practice.
In problems where the minimum arc length

a = min aij
(i,j)∈A

is greater than 1, the performance of the algorithm can be improved by

using a device suggested by Denardo and Fox [1979]. The idea is that the
label of a node cannot be reduced below b + a while searching bucket b,
so that no new nodes will be added to buckets b + 1, . . . , b + a − 1 while
searching bucket b. As a result, buckets b, b + 1, . . . , b + a − 1 can be lumped
into a single bucket. To take advantage of this idea, we can use
 
(N − 1)C + 1
a

buckets, and follow the strategy of placing node i into bucket b if

ab ≤ di ≤ a(b + 1) − 1.
 
The running time of the algorithm is then reduced to O A + (N C/a) .
72 The Shortest Path Problem Chap. 2

Bucket b 0 1 2 3 4 5 6 7 8

Contents of b 1 – 3,4,5 2,7 – 6 – – –

F IRST (b) 1 0 3 2 0 6 0 0 0

Node i 1 2 3 4 5 6 7

Label di 0 3 2 2 2 5 3

NEXT (i ) 0 7 4 5 0 0 0

PREVIOUS (i ) 0 0 0 3 4 0 2

Figure 2.8: Illustration of a doubly linked list data structure to maintain the can-
didate list V in buckets. In this example, the nodes in V are numbered 1, 2, . . . , 7,
and the buckets are numbered 0, 1, . . . , 8. A node i belongs to bucket b if di = b.
As shown in the first table, for each bucket b we maintain the first node of
the bucket in an array element F IRST (b), where F IRST (b) = 0 if bucket b is
empty.
As shown in the second table, for every node i we maintain two array
elements, N EXT (i) and P REV IOU S(i), giving the next node and the pre-
ceding node, respectively, of node i in the bucket where i is currently residing
[N EXT (i) = 0 or P REV IOU S(i) = 0 if i is the last node or the first node in its
bucket, respectively].

Another useful idea is that it is sufficient to maintain only C + 1

buckets, rather than (N − 1)C + 1, thereby significantly saving in memory.
The reason is that if we are currently searching bucket b, then all buckets
beyond b + C are known to be empty. To see this, note that the label dj of
any node j must be of the form di + aij , where i is a node that has already
been removed from the candidate list. Since di ≤ b and aij ≤ C, it follows
that dj ≤ b + C.
The idea of using buckets to maintain the nodes of the candidate
list can be
generalized considerably.
 In particular, buckets of width larger
than max 1, min(i,j)∈A aij may be used. This results in fewer buckets to
search over, thereby alleviating the O(N C) bottleneck of the running time
of the algorithm. There is a price for this, namely the need to search for a
minimum label node within the current bucket. This search can be speeded
up by using buckets with nonuniform widths, and by breaking down buckets
of large width into buckets of smaller width at the right moment. With
 Sec. 2.4 Label Correcting Methods 73

intelligent strategies of this type, one may obtain label setting methods
with very good polynomial complexity bounds; see Johnson [1977], Denardo
and Fox [1979], Ahuja, Mehlhorn, Orlin, and Tarjan [1990]. In practice,
however, the simpler algorithm of Dial has been more popular than these
methods.

2.4 LABEL CORRECTING METHODS

We now turn to the analysis of label correcting methods. In these methods,

the selection of the node to be removed from the candidate list V is simpler
and requires less overhead than in label setting methods, at the expense of
multiple entrances of nodes in V . All of these methods use some type of
queue to maintain the candidate list V . They differ in the way the queue
is structured, and in the choice of the queue position into which nodes
are inserted. In this section, we will discuss some of the most interesting
possibilities.

2.4.1 The Bellman-Ford Method

The simplest label correcting method uses a first-in first-out rule to update
the queue that is used to store the candidate list V . In particular, a node is
always removed from the top of the queue, and a node, upon entrance in the
candidate list, is placed at the bottom of the queue. Thus, it can be seen
that the method operates in cycles of iterations: the first cycle consists of
just iterating on node 1; in each subsequent cycle, the nodes that entered
the candidate list during the preceding cycle, are removed from the list
in the order that they were entered. We will refer to this method as the
Bellman-Ford method , because it is closely related to a method proposed
by Bellman [1957] and Ford [1956] based on dynamic programming ideas
(see Exercise 2.6).
The complexity analysis of the method is based on the following prop-
erty, which we will prove shortly:

Bellman-Ford Property
For each node i and integer k ≥ 1, let

dki = Shortest distance from 1 to i using paths that have k arcs or less,

where dki = ∞ if there is no path from 1 to i with k arcs or less. Then

the label di at the end of the kth cycle of iterations of the Bellman-Ford
method is less or equal to dki .
74 The Shortest Path Problem Chap. 2

In the case where all cycles have nonnegative length, the shortest
distance of every node can be achieved with a path having N − 1 arcs or
less, so the above Bellman-Ford property implies that the method finds
all the shortest distances after at most N − 1 cycles. Since each cycle
of iterations requires a total of O(A) operations (each arc is examined at
most once in each cycle), the running time of the Bellman-Ford method is
O(N A).
To prove the Bellman-Ford property, we first note that
 
dj = min dj , min {di + aij } ,
k+1 k k ∀ j, k ≥ 1, (2.10)
(i,j)∈A

since dk+1
j is either the length of a path from 1 to j with k arcs or less, in
which case it is equal to dkj , or else it is the length of some path that starts
at 1 goes to a predecessor node i with k arcs or less, and then goes to j
using arc (i, j). We now prove the Bellman-Ford property by induction.
At the end of the 1st cycle, we have for all i,
⎧
⎨0 if i = 1,
di = a1i if i = 1 and (1, i) ∈ A,
⎩
∞ if i = 1 and (1, i) ∈ / A,

while 
a1i if (1, i) ∈ A,
d1i =
∞ if (1, i) ∈
/ A,
so that di ≤ d1i for all i. Let di and V be the node labels and the contents
of the candidate list at the end of the kth cycle, respectively. Let also di be
the node labels at the end of the (k + 1)st cycle. We assume that di ≤ dki
for all i, and we will show that di ≤ dk+1i for all i. Indeed, by condition
(ii) of Prop. 2.2(a), we have

dj ≤ di + aij , ∀ (i, j) ∈ A with i ∈

/ V,

and since dj ≤ dj , it follows that

dj ≤ di + aij , ∀ (i, j) ∈ A with i ∈

/ V. (2.11)

We also have

dj ≤ di + aij , ∀ (i, j) ∈ A with i ∈ V, (2.12)

since at the time when i is removed from V , its current label, call it d˜i ,
satisfies d˜i ≤ di , and the label of j is set to d˜i + aij if it exceeds d˜i + aij .
By combining Eqs. (2.11) and (2.12), we see that

dj ≤ min {di + aij } ≤ min {dki + aij }, ∀ j, (2.13)

(i,j)∈A (i,j)∈A
Sec. 2.4 Label Correcting Methods 75

where the second inequality follows by the induction hypothesis. We also

have dj ≤ dj ≤ dkj by the induction hypothesis, so Eq. (2.13) yields
 
dj ≤ min dkj , min {dki + aij } = dk+1
j ,
(i,j)∈A

where the last equality holds by Eq. (2.10). This completes the induction
proof of the Bellman-Ford property.
The Bellman-Ford method can be used to detect the presence of a
negative cycle. Indeed, from Prop. 2.2, we see that the method fails to
terminate if and only if there exists a path that starts at 1 and contains
a negative cycle. Thus in view of the Bellman-Ford property, such a path
exists if and only if the algorithm has not terminated by the end of N − 1
cycles.
The best practical implementations of label correcting methods are
more sophisticated than the Bellman-Ford method. Their worst-case run-
ning time is no better than the O(N A) time of the Bellman-Ford method,
and in some cases it is considerably slower. Yet their practical performance
is often considerably better. We will discuss next three different types of
implementations.

2.4.2 The D’Esopo-Pape Algorithm

In this method, a node is always removed from the top of the queue used
to maintain the candidate list V . A node, upon entrance in the queue, is
placed at the bottom of the queue if it has never been in the queue before;
otherwise it is placed at the top.
The idea here is that when a node i is removed from the queue, its
label affects the labels of a subset Bi of the neighbor nodes j with (i, j) ∈ A.
When the label of i changes again, it is likely that the labels of the nodes
in Bi will require updating also. It is thus argued that it makes sense to
place the node at the top of the queue so that the labels of the nodes in Bi
get a chance to be updated as quickly as possible.
While this rationale is not quite convincing, it seems to work well in
practice for a broad variety of problems, including types of problems where
there are some negative arc lengths. On the other hand, special examples
have been constructed (Kershenbaum [1981], Shier and Witzgall [1981]),
where the D’Esopo-Pape algorithm performs very poorly. In particular, in
these examples, the number of entrances of some nodes in the candidate
list V is not polynomial. Computational studies have also shown that for
some classes of problems, the practical performance of the D’Esopo-Pape
algorithm can be very poor (Bertsekas [1993a]). Pallottino [1984], and
Gallo and Pallottino [1988] give a polynomial variant of the algorithm,
whose practical performance, however, is roughly similar to the one of the
original version.
76 The Shortest Path Problem Chap. 2

2.4.3 The SLF and LLL Algorithms

These methods are motivated by the hypothesis that when the arc lengths
are nonnegative, the queue management strategy should try to place nodes
with small labels near the top of the queue. For a supporting heuristic
argument, note that for a node j to reenter V , some node i such that
di + aij < dj must first exit V . Thus, the smaller dj was at the previous
exit of j from V the less likely it is that di +aij will subsequently become less
than dj for some node i ∈ V and arc (i, j). In particular, if dj ≤ mini∈V di
and the arc lengths aij are nonnegative, it is impossible that subsequent
to the exit of j from V we will have di + aij < dj for some i ∈ V .
We can think of Dijkstra’s method as implicitly placing at the top of
an imaginary queue the node with the smallest label, thereby resulting in
the minimal number N of iterations. The methods of this section attempt
to emulate approximately the minimum label selection policy of Dijkstra’s
algorithm with a much smaller computational overhead. They are primarily
suitable for the case of nonnegative arc lengths. While they will work even
when there are some negative arc lengths as per Prop. 2.2, there is no
reason to expect that in this case they will terminate faster (or slower)
than any of the other label correcting methods that we will discuss.
A simple strategy for placing nodes with small label near the top of the
queue is the Small Label First method (SLF for short). Here the candidate
list V is maintained as a double ended queue Q. At each iteration, the
node exiting V is the top node of Q. The rule for inserting new nodes is
given below:

SLF Strategy
Whenever a node j enters Q, its label dj is compared with the label
di of the top node i of Q. If dj ≤ di , node j is entered at the top of
Q; otherwise j is entered at the bottom of Q.

The SLF strategy provides a rule for inserting nodes in Q, but always
removes (selects for iteration) nodes from the top of Q. A more sophis-
ticated strategy is to make an effort to remove from Q nodes with small
labels. A simple possibility, called the Large Label Last method (LLL for
short) works as follows: At each iteration, when the node at the top of Q
has a larger label than the average node label in Q (defined as the sum of
the labels of the nodes in Q divided by the cardinality |Q| of Q), this node
is not removed from Q, but is instead repositioned to the bottom of Q.

LLL Strategy
Let i be the top node of Q, and let
Sec. 2.4 Label Correcting Methods 77

j∈Q dj
a= .
|Q|
If di > a, move i to the bottom of Q. Repeat until a node i such that
di ≤ a is found and is removed from Q.

It is simple to combine the SLF queue insertion and the LLL node
removal strategies, thereby obtaining a method referred to as SLF/LLL.
Experience suggests that, assuming nonnegative arc lengths, the SLF,
LLL, and combined SLF/LLL algorithms perform substantially faster than
the Bellman-Ford and the D’Esopo-Pape methods. The strategies are also
well-suited for parallel computation (see Bertsekas, Guerriero, and Mus-
manno [1996]). The combined SLF/LLL method consistently requires a
smaller number of iterations than either SLF or LLL, although the gain in
number of iterations is sometimes offset by the extra overhead.
Regarding the theoretical worst-case performance of the SLF and the
combined SLF/LLL algorithms, an example has been constructed by Chen
and Powell [1997], showing that these algorithms do not have polynomial
complexity in their pure form. However, nonpolynomial behavior seems
to be an extremely rare phenomenon in practice. In any case, one may
construct polynomial versions of the SLF and LLL algorithms, when the
arc lengths are nonnegative. A simple approach is to first sort the outgoing
arcs of each node by length. That is, when a node i is removed from Q, first
examine the outgoing arc from i that has minimum length, then examine
the arc of second minimum length, etc. This approach, due to Chen and
Powell [1997], can be shown to have complexity O(N A2 ) (see Exercise
2.9). Note, however, that sorting the outgoing arcs of a node by length
may involve significant overhead.
There is also another approach to construct polynomial versions of
the SLF and LLL algorithms (as well as other label correcting methods),
which leads to O(N A) complexity, assuming nonnegative arc lengths. To
see how this works, suppose that in the generic label correcting algorithm,
there is a set of increasing iteration indices t1 , t2 , . . . , tn+1 such that t1 = 1,
and for i = 1, . . . , n, all nodes that are in V at the start of iteration ti
are removed from V at least once prior to iteration ti+1 . Because all arc
lengths are nonnegative, this guarantees that the minimum label node of
V at the start of iteration ti will never reenter V after iteration ti+1 . Thus
the candidate list must have no more than N − i nodes at the start of
iteration ti+1 , and must become empty prior to iteration tN +1 . Thus, if
the running time of the algorithm between iterations ti and ti+1 is bounded
by R, the total running time of the algorithm will be bounded by N R, and
if R is polynomially bounded, the running time of the algorithm will also
be polynomially bounded.
Specializing now to the SLF and LLL cases, assume that between
78 The Shortest Path Problem Chap. 2

iterations ti and ti+1 , each node is inserted at the top of Q for a number
of times that is bounded by a constant and that (in the case of SLF/LLL)
the total number of repositionings is bounded by a constant multiple of
A. Then it can be seen that the running time of the algorithm between
iterations ti and ti+1 is O(A), and therefore the complexity of the algorithm
is O(N A).
To modify SLF or SLF/LLL so that they have an O(N A) worst-case
complexity, based on the preceding result, it is sufficient that we fix an inte-
ger k > 1, and that we separate the iterations of the algorithm in successive
blocks of kN iterations each. We then impose an additional restriction that,
within each block of kN iterations, each node can be inserted at most k − 1
times at the top of Q [that is, after the (k − 1)th insertion of a node to the
top of Q within a given block of kN iterations, all subsequent insertions of
that node within that block of kN iterations must be at the bottom of Q].
In the case of SLF/LLL, we also impose the additional restriction that the
total number of repositionings within each block of kN iterations should
be at most kA (that is, once the maximum number of kA repositionings is
reached, the top node of Q is removed from Q regardless of the value of its
label). The worst-case running times of the modified algorithms are then
O(N A). In practice, it is highly unlikely that the restrictions introduced
into the algorithms to guarantee O(N A) complexity will ever be exercised
if k is larger than a small number such as 3 or 4.

2.4.4 The Threshold Algorithm

Similar to the SLF/LLL methods, the premise of this algorithm is also

that, for nonnegative arc lengths, the number of iterations is reduced by
removing from the candidate list V nodes with relatively small label. In
the threshold algorithm, V is organized into two distinct queues Q
and Q


using a threshold parameter s. The queue Q
contains nodes with “small”
labels; that is, it contains only nodes whose labels are no larger than s.
At each iteration, a node is removed from Q
, and any node j to be added
to the candidate list is inserted at the bottom of Q
or Q

depending on
whether dj ≤ s or dj > s, respectively. When the queue Q
is exhausted,
the entire candidate list is repartitioned. The threshold is adjusted, and
the queues Q
and Q

are recalculated, so that Q
consists of the nodes
with labels that are no larger than the new threshold.
To understand how the threshold algorithm works, consider the case
of nonnegative arc lengths, and suppose that at time t the candidate list
is repartitioned based on a new threshold value s, and that at some sub-
sequent time t
> t the queue Q
gets exhausted. Then at time t
, all the
nodes of the candidate list have label greater than s. In view of the nonneg-
ativity of the arc lengths, this implies that all nodes with label less than or
equal to s will not reenter the candidate list after time t
. In particular, all
nodes that exited the candidate list between times t and t
become perma-
Sec. 2.4 Label Correcting Methods 79

nently labeled at time t
and never reenter the candidate list. We may thus
interpret the threshold algorithm as a block version of Dijkstra’s method ,
whereby a whole subset of nodes becomes permanently labeled when the
queue Q
gets exhausted.
The preceding interpretation suggests that the threshold algorithm is
suitable primarily for the case of nonnegative arc lengths (even though it
will work in general). Furthermore, the performance of the algorithm is
quite sensitive to the method used to adjust the threshold. For example, if
s is taken to be equal to the current minimum label, the method is identical
to Dijkstra’s algorithm; if s is larger than all node labels, Q

is empty and
the algorithm reduces to the generic label correcting method. With an
effective choice of threshold, the practical performance of the algorithm
is very good. A number of heuristic approaches have been developed for
selecting the threshold (see Glover, Klingman, and Phillips [1985], and
Glover, Klingman, Phillips, and Schneider [1985]). If all arc lengths are
nonnegative, a bound O(N A) on the operation count of the algorithm can
be shown; see Exercise 2.8(c).

Combinations of the Threshold and the SLF/LLL Methods

We mentioned earlier that the threshold algorithm may be interpreted as

a block version of Dijkstra’s method, whereby attention is restricted to the
subset of nodes that belong to the queue Q
, until this subset becomes per-
manently labeled. The algorithm used to permanently label the nodes of Q

is essentially the Bellman-Ford algorithm restricted to the subgraph defined
by Q
. It is possible to use a different algorithm for this purpose, based for
example on the SLF and LLL strategies. This motivates combinations of
the threshold and the SLF/LLL algorithms.
In particular, the LLL strategy can be used when selecting a node
to exit the queue Q
in the threshold algorithm (the top node of Q
is
repositioned to the bottom of Q
if its label is found smaller than the
average label in Q
). Furthermore, whenever a node enters the queue Q
,
it is added, according to the SLF strategy, at the bottom or the top of Q

depending on whether its label is greater than the label of the top node of
Q
or not. The same policy is used when transferring to Q
the nodes of
Q

whose labels do not exceed the current threshold parameter. Thus the
nodes of Q

are transferred to Q
one-by-one, and they are added to the
top or the bottom of Q
according to the SLF strategy. Finally, the SLF
strategy is also followed when a node enters the queue Q

.
Generally, the threshold strategy and the SLF/LLL strategy are com-
plementary and work synergistically. Computational experience suggests
that their combination performs extremely well in practice, and typically
results in an average number of iterations per node that is only slightly
larger than the minimum of 1 achieved by Dijkstra’s method. At the same
80 The Shortest Path Problem Chap. 2

time, these combined methods require considerably less overhead than Di-
jkstra’s method.

2.4.5 Comparison of Label Setting and Label Correcting

Let us now try to compare the two major special cases of the generic
algorithm, label setting and label correcting methods, assuming that the
arc lengths are nonnegative.
We mentioned earlier that label setting methods offer a better guar-
antee of good performance than label correcting methods, because their
worst-case running time is more favorable. In practice, however, there
are several considerations that argue in favor of label correcting methods.
One such consideration is that label correcting methods, because of their
inherent flexibility, are better suited for exploiting advanced initialization.
Another consideration is that when the graph is acyclic, label cor-
recting methods can be adapted to exploit the problem’s structure, so that
each node enters and exits the candidate list only once, thereby nullifying
the major advantage of label setting methods (see Exercise 2.10). The cor-
responding running time is O(A), which is the minimum possible. Note
that an important class of problems involving an acyclic graph is dynamic
programming (cf. Fig. 2.1).
A third consideration is that in practice, the graphs of shortest path
problems are often sparse; that is, the number of arcs is much smaller
than the maximum possible N 2 . In this case, efficient label correcting
methods tend to have a faster practical running time than label setting
methods. To understand the reason, note that all shortest path methods
require the unavoidable O(A) operations needed to scan once every arc, plus
some additional time which we can view as “overhead.” The overhead of
the popular label setting methods is roughly proportional to N in practice
(perhaps times a slowly growing factor, like log N ), as argued earlier for the
binary heap method and Dial’s algorithm. On the other hand, the overhead
of label correcting methods grows linearly with A (times a factor that likely
grows slowly), because for the most popular methods, the average number
of node entrances in the queue per node is typically not much larger than
1. Thus, we may conclude that the overhead ratio of label correcting to
label setting methods is roughly

A
· (a constant factor).
N

The constant factor above depends on the particular method used and
may vary slowly with the problem size, but is typically much less than 1.
Thus, the overhead ratio favors label correcting methods for a sparse graph
(A << N 2 ), and label setting methods for a dense graph (A ≈ N 2 ). This
is consistent with empirical observations.
 Sec. 2.5 Single Origin/Single Destination Methods 81

Let us finally note that label setting methods can take better advan-
tage of situations where only a small subset of the nodes are destinations, as
will be seen in the next section. This is also true of the auction algorithms
to be discussed in Section 2.6.

2.5 SINGLE ORIGIN/SINGLE DESTINATION METHODS

In this section, we discuss the adaptation of our earlier single origin/all

destination algorithms to the case where there is only one destination, call
it t, and we want to find the shortest distance from the origin node 1 to
t. We could of course use our earlier all-destinations algorithms, but some
improvements are possible.

2.5.1 Label Setting

Suppose that we use the label setting method. Then we can stop the
method when the destination t becomes permanently labeled; further com-
putation will not improve the label dt (Exercise 2.13 sharpens this criterion
in the case where min{i|(i,t)∈A} aij > 0). If t is closer to the origin than
many other nodes, the saving in computation time will be significant. Note
that this approach can also be used when there are several destinations.
The method is stopped when all destinations have become permanently
labeled.
Another possibility is to use a two-sided label setting method ; that is,
a method that simultaneously proceeds from the origin to the destination
and from the destination to the origin. In this method, we successively label
permanently the closest nodes to the origin (with their shortest distance
from the origin) and the closest nodes to the destination (with their shortest
distance to the destination). It can be shown that when some node gets
permanently labeled from both sides, the labeling can stop; by combining
the forward and backward paths of each labeled node and by comparing
the resulting origin-to-destination paths, one can obtain a shortest path.
Exercise 2.14 develops in some detail this approach, which can often lead
to a dramatic reduction in the total number of iterations. However, the
approach does not work when there are multiple destinations.

2.5.2 Label Correcting

Unfortunately, when label correcting methods are used, it may not be easy
to realize the savings just discussed in connection with label setting. The
difficulty is that even after we discover several paths to the destination t
(each marked by an entrance of t into V ), we cannot be sure that better
paths will not be discovered later. In the presence of additional problem
82 The Shortest Path Problem Chap. 2

structure, however, the number of times various nodes will enter V can be
reduced considerably, as we now explain.
Suppose that at the start of the algorithm we have, for each node i, an
underestimate ui of the shortest distance from i to t (we require ut = 0).
For example, if all arc lengths are nonnegative we may take ui = 0 for
all i. (We do not exclude the possibility that ui = −∞ for some i, which
corresponds to the case where no underestimate is available for the shortest
distance of i.) The following is a modified version of the generic shortest
path algorithm.
Initially
V = {1},
d1 = 0, di = ∞, ∀ i = 1.
The algorithm proceeds in iterations and terminates when V is empty. The
typical iteration (assuming V is nonempty) is as follows.

Iteration of the Generic Single Origin/Single Destination Al-

gorithm

Remove a node i from V . For each outgoing arc (i, j) ∈ A, if

di + aij < min{dj , dt − uj },

set
dj := di + aij
and add j to V if it does not already belong to V .

The preceding iteration is the same as the one of the all-destinations

generic algorithm, except that the test di + aij < dj for entering a node j
into V is replaced by the more stringent test di + aij < min{dj , dt − uj }.
(In fact, when the trivial underestimate uj = −∞ is used for all j = t the
two iterations coincide.) To understand the idea behind the iteration, note
that the label dj corresponds at all times to the best path found thus far
from 1 to j (cf. Prop. 2.2). Intuitively, the purpose of entering node j in
V when its label is reduced is to generate shorter paths to the destination
that pass through node j. If Pj is the path from 1 to j corresponding to
di + aij , then di + aij + uj is an underestimate of the shortest path length
among the collection of paths Pj that first follow path Pj to node j and
then follow some other path from j to t. However, if

di + aij + uj ≥ dt ,

the current best path to t, which corresponds to dt , is at least as short as

any of the paths in the collection Pj , which have Pj as their first component.
Sec. 2.5 Single Origin/Single Destination Methods 83

2 0
Origin 2 Destination

1 1 1 5
4 0
1
3 1

Iter. # Candidate List V Node Labels Node out of V

1 {1} (0, ∞, ∞, ∞, ∞) 1
2 {2, 3} (0, 2, 1, ∞, ∞) 2
3 {3, 5} (0, 2, 1, ∞, 2) 3
4 {5} (0, 2, 1, ∞, 2) 5
Ø (0, 2, 1, ∞, 2)

Figure 2.9: Illustration of the generic single origin/single destination algorithm.

Here the destination is t = 5 and the underestimates of shortest distances to t are
ui = 0 for all i. Note that at iteration 3, when node 3 is removed from V , the
label of node 4 is not improved to d4 = 2 and node 4 is not entered in V . The
reason is that d3 + a34 (which is equal to 2) is not smaller than d5 − u4 (which is
also equal to 2). Note also that upon termination the label of a node other than
t may not be equal to its shortest distance (e.g. d4 ).

It is unnecessary to consider such paths, and for this reason node j need
not be entered in V . In this way, the number of node entrances in V may
be sharply reduced.
Figure 2.9 illustrates the algorithm. The following proposition proves
its validity.

Proposition 2.4: Consider the generic single origin/single destina-

tion algorithm.
(a) At the end of each iteration, if dj < ∞, then dj is the length of
some path that starts at 1 and ends at j.
(b) If the algorithm terminates, then upon termination, either dt <
∞, in which case dt is the shortest distance from 1 to t, or else
there is no path from 1 to t.
(c) If the algorithm does not terminate, there exist paths of arbi-
trarily small length that start at 1.
84 The Shortest Path Problem Chap. 2

Proof: (a) The proof is identical to the corresponding part of Prop. 2.2.
(b) If upon termination we have dt = ∞, then the extra test di + aij + uj <
dt for entering V is always passed, so the algorithm generates the same
label sequences as the generic (all destinations) shortest path algorithm.
Therefore, Prop. 2.2(b) applies and shows that there is no path from 1 to t.
It will thus be sufficient to prove this part assuming that we have dt < ∞
upon termination.
Let dj be the final values of the labels dj obtained upon termination
and suppose that dt < ∞. Assume, to arrive at a contradiction, that there
is a path Pt = (1, j1 , j2 , . . . , jk , t) that has length Lt with Lt < dt . For
m = 1, . . . , k, let Ljm be the length of the path Pm = (1, j1 , j2 , . . . , jm ).
Let us focus on the node jk preceding t on the path Pt . We claim that
Ljk < djk . Indeed, if this were not so, then jk must have been removed at
some iteration from V with a label djk satisfying djk ≤ Ljk . If dt is the
label of t at the start of that iteration, we would then have

djk + ajk t ≤ Ljk + ajk t = Lt < dt ≤ dt ,

implying that the label of t would be reduced at that iteration from dt to

djk + ajk t , which is less than the final label dt – a contradiction.
Next we focus on the node jk−1 preceding jk and t on the path Pt . We
use a similar (though not identical) argument to show that Ljk−1 < djk−1 .
Indeed, if this were not so, then jk−1 must have been removed at some
iteration from V with a label djk−1 satisfying djk−1 ≤ Ljk−1 . If djk and dt
are the labels of jk and t at the start of that iteration, we would then have

djk−1 + ajk−1 jk ≤ Ljk−1 + ajk−1 jk = Ljk < djk ≤ djk ,

and since Ljk + ujk ≤ Lt < dt ≤ dt , we would also have

djk−1 + ajk−1 jk < dt − ujk .

From the above two equations, it follows that the label of jk would be
reduced at that iteration from djk to djk−1 + ajk−1 t , which is less than the
final label djk – a contradiction.
Proceeding similarly, we obtain Ljm < djm for all m = 1, . . . , k, and
in particular a1j1 = Lj1 < dj1 . Since

a1j1 + uj1 ≤ Lt < dt ,

and dt is monotonically nonincreasing throughout the algorithm, we see

that at the first iteration we will have a1j1 < min{dj1 , dt − uj1 }, so j1 will
enter V with the label a1j1 , which cannot be less than the final label dj1 .
This is a contradiction; the proof of part (b) is complete.
(c) The proof is identical to the proof of Prop. 2.2(c). Q.E.D.
Sec. 2.5 Single Origin/Single Destination Methods 85

There are a number of possible implementations of the algorithm of

this subsection, which parallel the ones given earlier for the many destina-
tions problem. An interesting possibility to speed up the algorithm arises
when an overestimate vj of the shortest distance from j to t is known a
priori . (We require that vt = 0. Furthermore, we set vj = ∞ if no overes-
timate is known for j.) The idea is that the method still works if the test
di + aij < dt − uj is replaced by the possibly sharper test di + aij < D − uj ,
where D is any overestimate of the shortest distance from 1 to t with D ≤ dt
(check the proof of Prop. 2.4). We can obtain estimates D that may be
strictly smaller than dt by using the scalars vj as follows: each time the
label of a node j is reduced, we check whether dj + vj < D; if this is so, we
replace D by dj + vj . In this way, we make the test for future admissibility
into the candidate list V more stringent and save some unnecessary node
entrances in V .

Advanced Initialization

We finally note that similar to the all-destinations case, the generic sin-
gle origin/single destination method need not be started with the initial
conditions
V = {1}, d1 = 0, di = ∞, ∀ i = 1.
The algorithm works correctly using several other initial conditions. One
possibility is to use for each node i, an initial label di that is either ∞ or
else it is the length of a path from 1 to i, and to take V = {i | di < ∞}.
A more sophisticated alternative is to initialize V so that it contains all
nodes i such that
di + aij < min{dj , dt − uj } for some (i, j) ∈ A.
This kind of initialization can be extremely useful when a “good”
path
P = (1, i1 , . . . , ik , t)
from 1 to t is known or can be found heuristically, and the arc lengths are
nonnegative so that we can use the underestimate ui = 0 for all i. Then
we can initialize the algorithm with

Length of portion of path P from 1 to i if i ∈ P ,
di =
∞ if i ∈
/ P,
V = {1, i1 , . . . , ik }.
If P is a near-optimal path and consequently the initial value dt is near its
final value, the test for future admissibility into the candidate list V will
be relatively tight from the start of the algorithm and many unnecessary
entrances of nodes into V may be saved. In particular, it can be seen that
all nodes whose shortest distances from the origin are greater or equal to
the length of P will never enter the candidate list.
 86 The Shortest Path Problem Chap. 2

2.6 AUCTION ALGORITHMS

In this section, we discuss another class of algorithms for finding a shortest

path from an origin s to a destination t. These are called auction algorithms
because they can be shown to be closely related to the naive auction algo-
rithm for the assignment problem discussed in Section 1.3 (see Bertsekas
[1991a], Section 4.3.3, or Bertsekas [1991b]). The main algorithm is very
simple. It maintains a single path starting at the origin. At each iteration,
the path is either extended by adding a new node, or contracted by deleting
its terminal node. When the destination becomes the terminal node of the
path, the algorithm terminates.
To get an intuitive sense of the algorithm, think of a mouse moving in
a graph-like maze, trying to reach a destination. The mouse criss-crosses
the maze, either advancing or backtracking along its current path. Each
time the mouse backtracks from a node, it records a measure of the desir-
ability of revisiting and advancing from that node in the future (this will
be represented by a suitable variable). The mouse revisits and proceeds
forward from a node when the node’s measure of desirability is judged
superior to those of other nodes. The algorithm emulates efficiently this
search process using simple data structures.  
The algorithm maintains a path P = (s, i1 ), (i1 , i2 ), . . . , (ik−1 , ik )
with no cycles, and modifies P using two operations, extension and con-
traction. If ik+1 is a node not on P and (ik , ik+1 ) is an arc, an extension of 
P by ik+1 replaces P by the path (s, i1 ), (i1 , i2 ), . . . , (ik−1 , ik ), (ik , ik+1 ) .
If P does not consist
 of just the origin node s, a contraction of P replaces
P by the path (s, i1 ), (i1 , i2 ), . . . , (ik−2 , ik−1 ) .
We introduce a variable pi for each node i, called the price of node i.
We denote by p the price vector consisting of all node prices. The algorithm
maintains a price vector p satisfying together with P the following property

pi ≤ aij + pj , for all arcs (i, j), (2.14)

pi = aij + pj , for all arcs (i, j) of P . (2.15)

If we view the prices pi as the negative of the labels di that we used earlier,
we see that the above conditions are equivalent to the CS conditions (2.2)
and (2.4). Consequently, we will also refer to Eqs. (2.14) and (2.15) as the
CS conditions. We assume that the initial pair (P, p) satisfies CS. This is
not restrictive, since the default pair

P = (s), pi = 0, for all i

satisfies CS in view of the nonnegative arc length assumption. To define

the algorithm we also need to assume that all cycles have positive length;
Exercise 2.17 indicates how this assumption can be relaxed.
Sec. 2.6 Auction Algorithms 87

It can be shown that if a pair (P, p) satisfies the CS conditions, then

the portion of P between node s and any node i ∈ P is a shortest path
from s to i, while ps −pi is the corresponding shortest distance. To see this,
note that by Eq. (2.15), pi − pk is the length of the portion of P between
i and k, and that every path connecting i to k must have length at least
equal to pi − pk [add Eq. (2.14) along the arcs of the path].
The algorithm proceeds in iterations, transforming a pair (P, p) sat-
isfying CS into another pair satisfying CS. At each iteration, the path P
is either extended by a new node or is contracted by deleting its terminal
node. In the latter case the price of the terminal node is increased strictly.
A degenerate case occurs when the path consists of just the origin node s;
in this case the path is either extended or is left unchanged with the price
ps being strictly increased. The iteration is as follows.

Iteration of the Auction Algorithm

Let i be the terminal node of P . If


pi < min aij + pj ,
{j|(i,j)∈A}

go to Step 1; else go to Step 2.

Step 1 (Contract path): Set


pi := min aij + pj ,
{j|(i,j)∈A}

and if i = s, contract P . Go to the next iteration.

Step 2 (Extend path): Extend P by node ji where


ji = arg min aij + pj
{j|(i,j)∈A}

(ties are broken arbitrarily). If ji is the destination t, stop; P is the

desired shortest path. Otherwise, go to the next iteration.

It is easily seen that the algorithm maintains CS. Furthermore, the

addition of the node ji to P following an extension does not create a cycle,
since otherwise, in view of the condition pi ≤ aij + pj , for every arc (i, j)
of the cycle we would have pi = aij + pj . By adding this equality along
the cycle, we see that the length of the cycle must be zero, which is not
possible by our assumptions.
Figure 2.10 illustrates the algorithm. It can be seen from the exam-
ple of this figure that the terminal node traces the tree of shortest paths
from the origin to the nodes that are closer to the origin than the given
88 The Shortest Path Problem Chap. 2

destination. This behavior is typical when the initial prices are all zero (see
Exercise 2.19).

p2 = 1.5

2 p1 = 2.5 2
1 1.5 p4 = 0

1 4 1 4
Origin Destination
2 3
3 3

p3 = 3

Shortest path problem with arc Trajectory of terminal node

lengths as shown and final prices generated by
the algorithm

Iteration # Path P prior Price vector p prior Type of action

to iteration to iteration during iteration

1 (1) (0, 0, 0, 0) contraction at 1

2 (1) (1, 0, 0, 0) extension to 2
3 (1, 2) (1, 0, 0, 0) contraction at 2
4 (1) (1, 1.5, 0, 0) contraction at 1
5 (1) (2, 1.5, 0, 0) extension to 3
6 (1, 3) (2, 1.5, 0, 0) contraction at 3
7 (1) (2, 1.5, 3, 0) contraction at 1
8 (1) (2.5, 1.5, 3, 0) extension to 2
9 (1, 2) (2.5, 1.5, 3, 0) extension to 4
10 (1, 2, 4) (2.5, 1.5, 3, 0) stop

Figure 2.10: An example illustrating the auction algorithm starting with P = (1)
and p = 0.

There is an interesting interpretation of the CS conditions in terms of

a mechanical model, due to Minty [1957]. Think of each node as a ball, and
for every arc (i, j), connect i and j with a string of length aij . (This requires
that aij = aji > 0, which we assume for the sake of the interpretation.) Let
the resulting balls-and-strings model be at an arbitrary position in three-
Sec. 2.6 Auction Algorithms 89

dimensional space, and let pi be the vertical coordinate of node i. Then

the CS condition pi − pj ≤ aij clearly holds for all arcs (i, j), as illustrated
in Fig. 2.11(b). If the model is picked up and left to hang from the origin
node (by gravity – strings that are tight are perfectly vertical), then for
all the tight strings (i, j) we have pi − pj = aij , so any tight chain of
strings corresponds to a shortest path between the end nodes of the chain,
as illustrated in Fig. 2.11(c). In particular, the length of the tight chain
connecting the origin node s to any other node i is ps − pi and is also equal
to the shortest distance from s to i.

2
1 1.5 Shortest path problem with
arc lengths shown next to the arcs.
1 4 Node 1 is the origin.
Node 4 is the destination.
2 3
3

(a)

p1 = 2.5 1

1
p1 1
1 2
2 p2 = 1.5
p2 2 2

p3 1.5
3
p3 = 0.5 3
p4 1.5 3
4
3
p4 = 0 4

(b) (c)

Figure 2.11: Illustration of the CS conditions for the shortest path problem. If
each node is a ball, and for every arc (i, j), nodes i and j are connected with a
string of length aij , the vertical coordinates pi of the nodes satisfy pi − pj ≤ aij ,
as shown in (b) for the problem given in (a). If the model is picked up and left
to hang from the origin node s, then ps − pi gives the shortest distance to each
node i, as shown in (c).

The algorithm can also be interpreted in terms of the balls-and-strings

model; it can be viewed as a process whereby nodes are raised in stages as
illustrated in Fig. 2.12. Initially all nodes are resting on a flat surface. At
each stage, we raise the last node in a tight chain that starts at the origin
to the level at which at least one more string becomes tight.
The following proposition establishes the validity of the auction algo-
rithm.
90 The Shortest Path Problem Chap. 2

2
1 1.5
Shortest path problem with
arc lengths shown next to the arcs.
1 4
Node 1 is the origin.
2 3 Node 4 is the destination.
3

(a)

2.0
1.5 2

1.0 1
1
0.5
1 2 3 4 2 3 4 3 4
0

Initial position After 1st stage After 2nd stage

3.0 3 3
2.5 1
2.0 1 1
1.5 2 2 2
1.0
0.5
3 4 4 4
0

After 3rd stage After 4th stage After 5th stage

The ball marked by gray (the terminal node of the

current path P) is raised at each stage.

(b)

Figure 2.12: Illustration of the auction algorithm in terms of the balls-and-

strings model for the problem shown in (a). The model initially rests on a flat
surface, and various balls are then raised in stages. At each stage we raise a
single ball i
= t (marked by gray), which is at a lower level than the origin s
and can be reached from s through a sequence of tight strings; i should not have
any tight string connecting it to another ball, which is at a lower level, that is, i
should be the last ball in a tight chain hanging from s. (If s does not have any
tight string connecting it to another ball, which is at a lower level, we use i = s.)
We then raise i to the first level at which one of the strings connecting it to a
ball at a lower level becomes tight. Each stage corresponds to a contraction plus
all the extensions up to the next contraction. The ball i, which is being raised,
corresponds to the terminal node of the current path P .
Sec. 2.6 Auction Algorithms 91

Proposition 2.5: If there exists at least one path from the origin
to the destination, the auction algorithm terminates with a shortest
path from the origin to the destination. Otherwise the algorithm never
terminates and ps → ∞.

Proof: We first show by induction that (P, p) satisfies the CS conditions

pi ≤ aij + pj , for all arcs (i, j), (2.16)

pi = aij + pj , for all arcs (i, j) of P , (2.17)

throughout the algorithm. Indeed, the initial pair satisfies CS by assump-
tion. Consider an iteration that starts with a pair (P, p) satisfying CS and
produces a pair (P , p). Let i be the terminal node of P . If


pi = aiji + pji = min aij + pj , (2.18)
{j|(i,j)∈A}

then P is the extension of P by the node ji and p = p, implying that the

CS condition (2.17) holds for all arcs of P as well as arc (i, ji ) [since ji
attains the minimum in Eq. (2.18)].
Suppose next that


pi < min aij + pj .
{j|(i,j)∈A}

Then if P is the degenerate path (s), the CS conditions hold vacuously.

Otherwise, P is obtained by contracting P , and for all nodes j ∈ P , we
have pj = pj , implying the CS conditions (2.16) and (2.17) for arcs outgoing
from nodes of P . Also, for the terminal node i, we have


pi = min aij + pj ,
{j|(i,j)∈A}

implying the CS condition (2.16) for arcs outgoing from that node as well.
Finally, since pi > pi and pk = pk for all k = i, we have pk ≤ akj + pj for
all arcs (k, j) outgoing from nodes k ∈/ P . This completes the induction
proof that (P, p) satisfies CS throughout the algorithm.
Assume first that there is a path from node s to the destination t.
By adding the CS condition (2.16) along that path, we see that ps − pt is
an underestimate of the (finite) shortest distance from s to t. Since ps is
monotonically nondecreasing, and pt is fixed throughout the algorithm, it
follows that ps must stay bounded.
We next claim that pi must stay bounded for all i. Indeed, in order to
have pi → ∞, node i must become the terminal node of P infinitely often.
92 The Shortest Path Problem Chap. 2

Each time this happens, ps − pi is equal to the shortest distance from s to

i, which is a contradiction since ps is bounded.
We next show that the algorithm terminates. Indeed, it can be seen
with a straightforward induction argument that for every node i, pi is either
equal to its initial value, or else it is the length of some path starting at i
plus the initial price of the final node of the path; we call this the modified
length of the path. Every path from s to i can be decomposed into a path
with no cycles together with a finite number of cycles, each having positive
length by assumption, so the number of distinct modified path lengths
within any bounded interval is bounded. Now pi was shown earlier to be
bounded, and each time i becomes the terminal node by extension of the
path P , pi is strictly larger over the preceding time i became the terminal
node of P , corresponding to a strictly larger modified path length. It follows
that the number of times i can become a terminal node by extension of the
path P is bounded. Since the number of path contractions between two
consecutive path extensions is bounded by the number of nodes in the
graph, the number of iterations of the algorithm is bounded, implying that
the algorithm terminates.
Assume now that there is no path from node s to the destination.
Then, the algorithm will never terminate, so by the preceding argument,
some node i will become the terminal node by extension of the path P
infinitely often and pi → ∞. At the end of iterations where this happens,
ps − pi must be equal to the shortest distance from s to i, implying that
ps → ∞. Q.E.D.

Nonpolynomial Behavior and Graph Reduction

A drawback of the auction algorithm as described above is that its running

time can depend on the arc lengths. A typical situation arises in graphs
involving a cycle with relatively small length, as illustrated in Fig. 2.13.
It is possible to turn the algorithm into one that is polynomial, by using
some variations of the algorithm. In these variations, in addition to the
extension and contraction operations, an additional reduction operation is
introduced whereby some unnecessary arcs of the graph are deleted. We
briefly describe the simplest of these variations, and we refer to Bertsekas,
Pallottino, and Scutellà [1995] for other more sophisticated variations and
complexity analysis.
This variant of the auction algorithm has the following added feature:
each time that a node j becomes the terminal node of the path P through
an extension using arc (i, j), all incoming arcs (k, j) of j with k = i are
deleted from the graph. Also, each time that a node j with no outgoing
arcs becomes the terminal node of P , the path P is contracted and the
node j is deleted from the graph. It can be seen that the arc deletion
process leaves the shortest distance from s to t unaffected, and that the
Sec. 2.6 Auction Algorithms 93

Figure 2.13: Example graph for which

the number of iterations of the algo-
rithm is not polynomially bounded. The
Origin Destination lengths are shown next to the arcs and
5 L > 1. By tracing the steps of the algo-
1 2 3
1 1 L rithm starting with P = (1) and p = 0,
1 1 we see that the price of node 3 will be
4 first increased by 1 and then it will be
increased by increments of 3 (the length
of the cycle) as many times as is neces-
sary for p3 to reach or exceed L.

algorithm terminates either by finding a shortest path from s to t or by

deleting s, depending on whether there exists at least one path from s to t
or not. It can also be seen that this is also true even if there are cycles of
zero length. Thus, in addition to addressing the nonpolynomial behavior,
the graph reduction process deals effectively with the case where there are
zero length cycles.
As an illustration, the reader may apply the algorithm with graph
reduction to the example of Fig. 2.13. After the first iteration when node 2
becomes the terminal node of P for the first time, the arc (4, 2) is deleted,
and the cycle (2, 3, 4, 2) that caused the nonpolynomial behavior is elim-
inated. Furthermore, once node 4 becomes the terminal node of P , it
gets deleted because it no longer has any outgoing arcs. The number of
iterations required is greatly reduced.
The effect of graph reduction may be enhanced by introducing a fur-
ther idea due to Cerulli (see Cerulli, Festa, and Raiconi [1997a]). In par-
ticular, if in the process of eliminating arcs, a node i is left with only one
outgoing arc (i, j), it may be “combined” with node j. This can be done
efficiently, and may result in significant computational savings for some
problem types (particularly those involving a sparse graph).
In addition to graph reduction, there are a number of ideas that can be
used to implement efficiently the auction algorithm; see Bertsekas [1991b],
Bertsekas, Pallottino, and Scutellà [1995], and Cerulli, Festa, and Raiconi
[1997b].

The Case of Multiple Destinations or Multiple Origins

To solve the problem with multiple destinations and a single origin, one
can simply run the algorithm until every destination becomes the terminal
node of the path at least once. Also, to solve the problem with multiple
origins and a single destination, one can combine several versions of the
algorithm – one for each origin. However, the different versions can share a
common price vector, since regardless of the origin considered, the condition
pi ≤ aij + pj is always maintained. There are several ways to operate such
a method; they differ in the policy used for switching between different
94 The Shortest Path Problem Chap. 2

origins. One possibility is to run the algorithm for one origin and, after the
shortest path is obtained, to switch to the next origin (without changing the
price vector), and so on, until all origins are exhausted. Another possibility,
which is probably preferable in most cases, is to rotate between different
origins, switching from one origin to another, if a contraction at the origin
occurs or the destination becomes the terminal node of the current path.

The Reverse Algorithm

For problems with one origin and one destination, a two-sided version of
the algorithm is particularly effective. This method maintains, in addition
to the path P , another path R that ends at the destination. To understand
this version, we first note that in shortest path problems, one can exchange
the role of origins and destinations by reversing the direction of all arcs.
It is therefore possible to use a destination-oriented version of the auction
algorithm that maintains a path R that ends at the destination and changes
at each iteration by means of a contraction or an extension. This algorithm,
called the reverse algorithm, is mathematically equivalent to the earlier
(forward) auction algorithm. Initially, in the reverse algorithm, R is any
path ending at the destination, and p is any price vector satisfying CS
together with R; for example,

R = (t), pi = 0, for all i,

if all arc lengths are nonnegative.

Iteration of the Reverse Algorithm

Let j be the starting node of R. If


pj > max pi − aij ,
{i|(i,j)∈A}

go to Step 1; else go to Step 2.

Step 1: (Contract path) Set


pj := max pi − aij ,
{i|(i,j)∈A}

and if j = t, contract R, (that is, delete the starting node j of R). Go

to the next iteration.
Step 2: (Extend path) Extend R by node ij , (that is, make ij the
starting node of R, preceding j), where
Sec. 2.6 Auction Algorithms 95


ij = arg max pi − aij
{i|(i,j)∈A}

(ties are broken arbitrarily). If ij is the origin s, stop; R is the desired

shortest path. Otherwise, go to the next iteration.

The reverse algorithm is most helpful when it is combined with the

forward algorithm. In a combined algorithm, initially we have a price
vector p, and two paths P and R, satisfying CS together with p, where P
starts at the origin and R ends at the destination. The paths P and R
are extended and contracted according to the rules of the forward and the
reverse algorithms, respectively, and the combined algorithm terminates
when P and R have a common node. Both P and R satisfy CS together
with p throughout the algorithm, so when P and R meet, say at node i,
the composite path consisting of the portion of P from s to i followed by
the portion of R from i to t will be shortest.

Combined Forward/Reverse Auction Algorithm

Step 1: (Run forward algorithm) Execute several iterations of the

forward algorithm (subject to the termination condition), at least one
of which leads to an increase of the origin price ps . Go to Step 2.
Step 2: (Run reverse algorithm) Execute several iterations of the
reverse algorithm (subject to the termination condition), at least one
of which leads to a decrease of the destination price pt . Go to Step 1.

The combined forward/reverse algorithm can also be interpreted in

terms of the balls-and-strings model of Fig. 2.11. Again, all nodes are
resting initially on a flat surface. When the forward part of the algorithm
is used, we raise nodes in stages as illustrated in Fig. 2.12. When the
reverse part of the algorithm is used, we lower nodes in stages; at each
stage, we lower the top node in a tight chain that ends at the destination
to the level at which at least one more string becomes tight.
The combined forward/reverse auction algorithm can be easily ex-
tended to handle single-origin/many-destination problems. One may start
the reverse portion of the algorithm from any destination for which a
shortest path has not yet been found. Based on experiments with ran-
domly generated problems, the combined forward/reverse auction algo-
rithm (with graph reduction to eliminate nonpolynomial behavior) out-
performs substantially and often dramatically its closest competitors for
single-origin/few-destination problems (see Bertsekas [1991b], and Bert-
sekas, Pallottino, and Scutellà [1995]). The intuitive reason for this is that
through the mechanism of the reverse portion of the algorithm, the selected
 96 The Shortest Path Problem Chap. 2

destinations are reached by the forward portion faster than other nodes,
thereby leading to faster termination.

2.7 MULTIPLE ORIGIN/MULTIPLE DESTINATION METHODS

In this section, we consider the all-pairs shortest path problem, where we

want to find a shortest path from each node to each other node. The Floyd-
Warshall algorithm is specifically designed for this problem, and it is not
any faster when applied to the single destination problem. It starts with
the initial condition

0 aij if (i, j) ∈ A,
Dij =
∞ otherwise,

and generates sequentially for all k = 0, 1, . . . , N − 1, and all nodes i and

j,   
k+1
Dij = min D k , Dk
ij i(k+1) + D k
(k+1)j if j = i,
∞ otherwise.
An induction argument shows that Dij k gives the shortest distance

from node i to node j using only nodes from 1 to k as intermediate nodes.

Thus, DijN gives the shortest distance from i to j (with no restriction on

the intermediate nodes). There are N iterations, each requiring O(N 2 )

operations, for a total of O(N 3 ) operations.
Unfortunately, the Floyd-Warshall algorithm cannot take advantage
of sparsity of the graph. It appears that for sparse problems it is typically
better to apply a single origin/all destinations algorithm separately for each
origin. If all the arc lengths are nonnegative, a label setting method can
be used separately for each origin. If there are negative arc lengths (but no
negative length cycles), one can of course apply a label correcting method
separately for each origin, but there is another alternative that results in
a superior worst-case complexity. It is possible to apply a label correcting
method only once to a single origin/all destinations problem and obtain
an equivalent all-pairs shortest path problem with nonnegative arc lengths;
the latter problem can be solved using N separate applications of a label
setting method. This alternative is based on the following proposition,
which applies to the general minimum cost flow problem.

Proposition 2.7: Every minimum cost flow problem with arc costs
aij such that all simple forward cycles have nonnegative cost is equiv-
alent to another minimum cost flow problem involving the same graph
and nonnegative arc costs âij of the form
Sec. 2.7 Multiple Origin/Multiple Destination Methods 97

âij = aij + di − dj , ∀ (i, j) ∈ A,

where the scalars di can be found by solving a single origin/all des-
tinations shortest path problem. The two problems are equivalent in
the sense that they have the same constraints, and the cost function
of one is the same as the cost function of the other plus a constant.

Proof: Let (N , A) be the graph of the given problem. Introduce a new

node 0 and an arc (0, i) for each i ∈ N , thereby obtaining a new graph
(N
, A
). Consider the shortest path problem involving this graph, with
arc lengths aij for the arcs (i, j) ∈ A and 0 for the arcs (0, i). Since all
incident arcs of node 0 are outgoing, all simple forward cycles of (N
, A
) are
also simple forward cycles of (N , A) and, by assumption, have nonnegative
length. Since any forward cycle can be decomposed into a collection of
simple forward cycles (cf. Exercise 1.4 in Chapter 1), all forward cycles
(not necessarily simple) of (N
, A
) have nonnegative length. Furthermore,
there is at least one path from node 0 to every other node i, namely the
path consisting of arc (0, i). Therefore, the shortest distances di from node
0 to all other nodes i can be found by a label correcting method, and by
Prop. 2.2, we have

âij = aij + di − dj ≥ 0, ∀ (i, j) ∈ A.

Let us now view (i,j)∈A âij xij as the cost function of a minimum
cost flow problem involving the graph (N , A) and the constraints of the
original problem. We have
   
âij xij = aij + di − dj xij
(i,j)∈A (i,j)∈A
⎛ ⎞
   
= aij xij + di ⎝ xij − xji ⎠
(i,j)∈A i∈N {j|(i,j)∈A} {j|(j,i)∈A}
 
= aij xij + di si ,
(i,j)∈A i∈N

where si is the supply of node i. Thus, the two cost functions (i,j)∈A âij xij

and (i,j)∈A aij xij differ by the constant i∈N di si . Q.E.D.

It can be seen now that the all-pairs shortest path problem can be
solved by using a label correcting method to solve the single origin/all
destinations problem described in the above proof, thereby obtaining the
scalars di and
âij = aij + di − dj , ∀ (i, j) ∈ A,
 98 The Shortest Path Problem Chap. 2

and by then applying a label setting method N times to solve the all-
pairs shortest path problem involving the nonnegative arc lengths âij . The
shortest distance Dij from i to j is obtained by subtracting di −dj from the
shortest distance from i to j found by the label setting method. To estimate
the running time of this approach, note that the label correcting method
requires O(N A) computation using the Bellman-Ford method, and each
of the N applications of the label setting method require less than O(N 2 )
computation (the exact count depends on the method used). Thus the
overall running time is less that the O(N 3 ) required by the Floyd-Warshall
algorithm, at least for sparse graphs.
Still another possibility for solving the all-pairs shortest path problem
is to solve N separate single origin/all destinations problems but to also use
the results of the computation for one origin to start the computation for
the next origin; see our earlier discussion of initialization of label correcting
methods and also the discussion at the end of Section 5.2.

2.8 NOTES, SOURCES, AND EXERCISES

The work on the shortest path problem is very extensive, so we will re-
strict ourselves to citing the references that relate most to the material
presented. Literature surveys are given by Dreyfus [1969], Deo and Pang
[1984], and Gallo and Pallottino [1988]. The latter reference also contains
codes for the most popular shortest path methods, and extensive compu-
tational comparisons. A survey of applications in transportation networks
is given in Pallottino and Scutellà [1997a]. Parallel computation aspects of
shortest path algorithms, including asynchronous versions of some of the
algorithms developed here, are discussed in Bertsekas and Tsitsiklis [1989],
and Kumar, Grama, Gupta, and Karypis [1994].
The generic algorithm was proposed as a unifying framework of many
of the existing shortest path algorithms in Pallottino [1984], and Gallo
and Pallottino [1986]. The first label setting method was suggested in
Dijkstra [1959], and also independently in Dantzig [1960], and Whitting
and Hillier [1960]. The binary heap method was proposed by Johnson
[1972]. Dial’s algorithm (Dial [1969]) received considerable attention after
the appearance of the paper by Dial, Glover, Karney, and Klingman [1979];
see also Denardo and Fox [1979].
The Bellman-Ford algorithm was proposed in Bellman [1957] and
Ford [1956] in the form given in Exercise 2.6, where the labels of all nodes
are iterated simultaneously. The D’Esopo-Pape algorithm appeared in
Pape [1974] based on an earlier suggestion of D’Esopo. The SLF and
SLF/LLL methods were proposed by Bertsekas [1993a], and by Bertsekas,
Guerriero, and Musmanno [1996]. Chen and Powell [1997] gave a simple
polynomial version of the SLF method (Exercise 2.9). The threshold al-
Sec. 2.8 Notes, Sources, and Exercises 99

gorithm was developed by Glover, Klingman, and Phillips [1985], Glover,

Klingman, Phillips, and Schneider [1985], and Glover, Glover, and Kling-
man [1986].
Two-sided label setting methods for the single origin/single destina-
tion problem (Exercise 2.14) were proposed by Nicholson [1966]; see also
Helgason, Kennington, and Stewart [1993], which contains extensive com-
putational results. The idea of using underestimates of the shortest dis-
tance to the destination in label correcting methods originated with the A∗
algorithm, a shortest path algorithm that is popular in artificial intelligence
(see Nilsson [1971], [1980], and Pearl [1984]).
The Floyd-Warshall algorithm was given in Floyd [1962] and uses a
theorem due to Warshall [1962]. Alternative algorithms for the all-pairs
problem are given in Dantzig [1967] and Tabourier [1973]. Reoptimization
approaches that use the results of a shortest path computation for one
origin to initialize the computation for other origins are given by Gallo and
Pallottino [1982], and Florian, Nguyen, and Pallottino [1981].
The auction algorithm for shortest paths is due to Bertsekas [1991b].
The idea of graph reduction was proposed by Pallottino and Scutellà [1991],
and an O(N 3 ) implementation of an auction algorithm with graph reduc-
tion was given by Bertsekas, Pallottino, and Scutellà [1995]. An analysis
of a parallel asynchronous implementation is given by Polymenakos and
Bertsekas [1994]. Some variants of the auction algorithm that use slightly
different price updating schemes have been proposed in Cerulli, De Leone,
and Piacente [1992], and Bertsekas [1992b] (see Exercise 2.33). A method
that combines the auction algorithm with some dual price iterations was
given by Pallottino and Scutellà [1997b].

EXERCISES

2.1

Consider the graph of Fig. 2.14. Find a shortest path from 1 to all nodes using
the binary heap method, Dial’s algorithm, the D’Esopo-Pape algorithm, the SLF
method, and the SLF/LLL method.

2.2

Suppose that the only arcs that have negative lengths are outgoing from the
origin node 1. Show how to adapt Dijkstra’s algorithm so that it solves the
all-destinations shortest path problem in at most N − 1 iterations.
100 The Shortest Path Problem Chap. 2

1
2 4

4 5 2
0 Figure 2.14: Graph for Exercise
1 8 2.1. The arc lengths are the num-
6
bers shown next to the arcs.
9
1
0 5
3 5
5

2.3

Give an example of a problem where the generic shortest path algorithm will
reduce the label of node 1 to a negative value.

2.4 (Shortest Path Tree Construction)

Consider the single origin/all destinations shortest path problem and assume that
all cycles have nonnegative length. Consider the generic algorithm of Section 2.2,
and assume that each time a label dj is decreased to di +aij the arc (i, j) is stored
in an array PRED(j ). Consider the subgraph of the arcs PRED(j ), j ∈ N , j = 1.
Show that at the end of each iteration this subgraph is a tree rooted at the origin,
and that upon termination it is a tree of shortest paths.

2.5 (Uniqueness of Solution of Bellman’s Equation)

Assume that all cycles have positive length. Show that if the scalars d1 , d2 , . . . , dN
satisfy
dj = min {di + aij }, ∀ j = 1,
(i,j)∈A

d1 = 0,

then for all j, dj is the shortest distance from 1 to j. Show by example that this
need not be true if there is a cycle of length 0. Hint: Consider the arcs (i, j)
attaining the minimum in the above equation and consider the paths formed by
these arcs.

2.6 (The Original Bellman-Ford Method)

Consider the single origin/all destinations shortest path problem. The Bellman-
Ford method, as originally proposed by Bellman and Ford, updates the labels of
all nodes simultaneously in a single iteration. In particular, it starts with the
initial conditions
d01 = 0, d0j = ∞, ∀ j = 1,
Sec. 2.8 Notes, Sources, and Exercises 101

and generates dkj , k = 1, 2, . . ., according to

dk1 = 0, dkj = min {dk−1

i + aij }, ∀ j = 1.
(i,j)∈A

(a) Show that for all j = 1 and k ≥ 1, dkj is the shortest distance from 1 to
j using paths with k arcs or less, where dkj = ∞ means that all the paths
from 1 to j have more than k arcs.
(b) Assume that all cycles have nonnegative length. Show that the algorithm
terminates after at most N iterations, in the sense that for some k ≤ N we
have dkj = dk−1
j for all j. Conclude that the running time of the algorithm
is O(N A).

2.7 (The Bellman-Ford Method with Arbitrary Initialization)

Consider the single origin/all destinations shortest path problem and the follow-
ing variant of the Bellman-Ford method of Exercise 2.6:

dk1 = 0, dkj = min {dk−1

i + aij }, ∀ j = 1,
(i,j)∈A

where each of the initial iterates d0i is an arbitrary scalar or ∞, except that
d01 = 0. We say that the algorithm terminates after k iterations if dki = dk−1
i for
all i.
(a) Given nodes i = 1 and j = 1, define
k
wij = minimum path length over all paths starting at i, ending at j,
k
and having k arcs (wij = ∞ if there is no such path).

For i = 1 and j = 1, define

k
w1j = minimum path length over all paths from 1 to j having k arcs or less
k
(w1j = ∞ if there is no such path).

Show by induction that

dkj = min {d0j + wij

k
}, ∀ j = 2, . . . , N, and k ≥ 1.
i=1,...,N

(b) Assume that there exists a path from 1 to every node i and that all cycles
have positive length. Show that the method terminates at some iteration
k, with dki equal to the shortest distances d∗i . Hint: For all i = 1 and j = 1,
limk→∞ wij k
= ∞, while for all j = 1, w1j
k
= d∗j for all k ≥ N − 1.
(c) Under the assumptions of part (b), show that if d0i ≥ d∗i for all i = 1, the
method terminates after at most m∗ + 1 iterations, where

m∗ = max mi ≤ N − 1,
i=1
102 The Shortest Path Problem Chap. 2

and mi is the smallest number of arcs contained in a shortest path from 1

to i.
(d) Under the assumptions of part (b), let
β = max{d∗i − d0i },
i=1

and assume that β > 0. Show that the method terminates after at most
k + 1 iterations, where k = N − 1 if the graph is acyclic, and k = N − 2 −
β/L if the graph has cycles, where
Length of the cycle
L= min ,
All simple cycles Number of arcs on the cycle
is the, so called, minimum cycle mean of the graph. Note: See Section 4.1 of
Bertsekas and Tsitsiklis [1989] for related analysis, and an example showing
that the given upper bound on the number of iterations for termination is
tight.
(e) (Finding the minimum cycle mean) Consider the following Bellman-Ford-
like algorithm:
dk (i) = min {aij + dk−1 (j)}, ∀ i = 1, . . . , N,
(i,j)∈A

d0 (i) = 0, ∀ i = 1, . . . , N.
We assume that there exists at least one cycle, but we do not assume that
all cycles have positive length. Show that the minimum cycle mean L of
part (d) is given by
dN (i) − dk (i)
L= min max .
i=1,...,N k=0,...,N −1 N −k
k
Hint: Show that d (i) is equal to the minimum path length over all paths
that start at i and have k arcs.

2.8 (Complexity of the Generic Algorithm)

Consider the generic algorithm, assuming that all arc lengths are nonnegative.
(a) Consider a node j satisfying at some time
dj ≤ di , ∀ i ∈ V.
Show that this relation will be satisfied at all subsequent times and that j
will never again enter V . Furthermore, dj will remain unchanged.
(b) Suppose that the algorithm is structured so that it removes from V a node
of minimum label at least once every k iterations (k is some integer). Show
that the algorithm will terminate in at most kN iterations.
(c) Show that the running time of the threshold algorithm is O(N A). Hint:
Define a cycle to be a sequence of iterations between successive repartition-
ings of the candidate list V . In each cycle, the node of V with minimum
label at the start of the cycle will be removed from V during the cycle.
Sec. 2.8 Notes, Sources, and Exercises 103

2.9 (Complexity of the SLF Method)

The purpose of this exercise, due to Chen and Powell [1997], is to show one way
to use the SLF method so that it has polynomial complexity. Suppose that the
outgoing arcs of each node have been presorted in increasing order by length.
The effect of this, in the context of the generic shortest path algorithm, is that
when a node i is removed from the candidate list, we first examine the outgoing
arc from i that has minimum length, then we examine the arc of second minimum
length, etc. Show an O(N A2 ) complexity bound for the method.

2.10 (Label Correcting for Acyclic Graphs)

Consider the problem of finding shortest paths from the origin node 1 to all
destinations, and assume that the graph does not contain any forward cycles.
Let Tk be the set of nodes i such that every path from 1 to i has k arcs or more,
and there exists a path from 1 to i with exactly k arcs. For each i, if i ∈ Tk define
IN DEX(i) = k. Consider a label setting method that selects a node i from the
candidate list that has minimum IN DEX(i).
(a) Show that the method terminates and that each node visits the candidate
list at most once.
(b) Show that the sets Tk can be constructed in O(A) time, and that the
running time of the algorithm is also O(A).

2.11

Consider the graph of Fig. 2.14. Find a shortest path from node 1 to node 6
using the generic single origin/single destination method of Section 2.5 with all
distance underestimates equal to zero.

2.12

Consider the problem of finding a shortest path from the origin 1 to a single des-
tination t, subject to the constraint that the path includes a given node s. Show
how to solve this problem using the single origin/single destination algorithms of
Section 2.5.

2.13 (Label Setting for Few Destinations)

Consider a label setting approach for finding shortest paths from the origin node
1 to a selected subset of destinations T . Let
a= min ait ,
{(i,t)∈A|t∈T }

and assume that a > 0. Show that one may stop the method when the node of
minimum label in V has a label dmin that satisfies
dmin + a ≥ max dt .
t∈T
104 The Shortest Path Problem Chap. 2

2.14 (Two-Sided Label Setting)

Consider the shortest path problem from an origin node 1 to a destination node
t, and assume that all arc lengths are nonnegative. This exercise considers an
algorithm where label setting is applied simultaneously and independently from
the origin and from the destination. In particular, the algorithm maintains a
subset of nodes W , which are permanently labeled from the origin, and a subset
of nodes V , which are permanently labeled from the destination. When W and
V have a node i in common the algorithm terminates. The idea is that a shortest
path from 1 to t cannot contain a node j ∈/ W ∪ V ; any such path must be longer
than a shortest path from 1 to i followed by a shortest path from i to t (unless j
and i are equally close to both 1 and to t).
Consider two subsets of nodes W and V with the following properties:
(1) 1 ∈ W and t ∈ V .
(2) W and V have nonempty intersection.
(3) If i ∈ W and j ∈/ W , then the shortest distance from 1 to i is less than or
equal to the shortest distance from 1 to j.
(4) If i ∈ V and j ∈/ V , then the shortest distance from i to t is less than or
equal to the shortest distance from j to t.
Let d1i be the shortest distance from 1 to i using paths all the nodes of which,
with the possible exception of i, lie in W (d1i = ∞ if no such path exists), and let
dti be the shortest distance from i to t using paths all the nodes of which, with
the possible exception of i, lie in V (dti = ∞ if no such path exists).
(a) Show that such W , V , d1i , and dti can be found by applying a label setting
method simultaneously for the single origin problem with origin node 1 and
for the single destination problem with destination node t.
(b) Show that the shortest distance D1t from 1 to t is given by




D1t = min d1i + dti = min d1i + dti = min d1i + dti .
i∈W i∈W ∪V i∈V

(c) Show that the nonempty

intersection
 condition (2) can be replaced by the
condition mini∈W d1i + dti ≤ maxi∈W d1i + maxi∈V dti .

2.15

Apply the forward/reverse auction algorithm to the example of Fig. 2.13, and
show that it terminates in a number of iterations that does not depend on the
large arc length L. Construct a related example for which the number of iterations
of the forward/reverse algorithm is not polynomially bounded.

2.16 (Finding an Initial Price Vector)

In order to initialize the auction algorithm, one needs a price vector p satisfying
the condition
pi ≤ aij + pj , ∀ (i, j) ∈ A. (2.19)
Sec. 2.8 Notes, Sources, and Exercises 105

Such a vector may not be available if some arc lengths are negative. Further-
more, even if all arc lengths are nonnegative, there are many cases where it is
important to use a favorable initial price vector in place of the default choice
p = 0. This possibility arises in a reoptimization context with slightly different
arc length data, or with a different origin and/or destination. This exercise gives
an algorithm to obtain a vector p satisfying the condition (2.19), starting from
another vector p satisfying the same condition for a different set of arc lengths
aij .
Suppose that we have a vector p̄ and a set of arc lengths {āij }, satisfying
pi ≤ aij + pj for all arcs (i, j), and we are given a new set of arc lengths {aij }.
(For the case where some arc lengths aij are negative, this situation arises with
p = 0 and aij = max{0, aij }.) Consider the following algorithm that maintains a
subset of arcs E and a price vector p, and terminates when E is empty. Initially

E = {(i, j) ∈ A | aij < āij , i = t}, p = p̄.

The typical iteration is as follows:

Step 1 (Select arc to scan): If E is empty, stop; otherwise, remove an arc
(i, j) from E and go to Step 2.
Step 2 (Add affected arcs to E): If pi > aij + pj , set

pi := aij + pj

and add to E every arc (k, i) with k = t that does not already belong to E.
Assuming that each node i is connected to the destination t with at least one
path, and that all cycle lengths are positive, show that the algorithm terminates
with a price vector p satisfying

pi ≤ aij + pj , ∀ (i, j) ∈ A with i = t.

2.17 (Extension for the Case of Zero Length Cycles)

Extend the auction algorithm for the case where all arcs have nonnegative length
but some cycles may consist exclusively of zero length arcs. Hint: Any cycle of
zero length arcs generated by the algorithm can be treated as a single node. An
alternative is the idea of graph reduction discussed in Section 2.6.

2.18

Consider the two single origin/single destination shortest path problems shown
in Fig. 2.15.
(a) Show that the number of iterations required by the forward auction algo-
rithm is estimated accurately by

nt − 1 + (2ni − 1),
i∈I, i=t
106 The Shortest Path Problem Chap. 2

where ni is the number of nodes in a shortest path from 1 to i. Show also

that the corresponding running times are O(N 2 ).
(b) Show that for the problem of Fig. 2.15(a) the running time of the for-
ward/reverse auction algorithm (with a suitable “reasonable” rule for switch-
ing between the forward and reverse algorithms) is O(N 2 ) (the number of
iterations is roughly half the corresponding number for the forward algo-
rithm). Show also that for the problem of Fig. 2.15(b) the running time of
the forward/reverse algorithm is O(N ).

1 2 3 . . . . N- 1 t (a)

3
1 t (b)
. . . .

N-1

Figure 2.15: Shortest path problems for Exercise 2.18. In problem (a) the arc
lengths are equal to 1. In problem (b), the length of each arc (1, i) is i, and the
length of each arc (i, t) is N .

2.19

In the auction algorithm of Section 2.6, let ki be the first iteration at which node
i becomes the terminal node of the path P . Show that if ki < kj , then the
shortest distance from 1 to i is less or equal to the shortest distance from 1 to j.

2.20 (A Forward/Reverse Version of Dijkstra’s Algorithm)

Consider the single origin/single destination shortest path problem and assume
that all arc lengths are nonnegative. Let node 1 be the origin, let node t be
the destination, and assume that there exists at least one path from 1 to t.
This exercise provides a forward/reverse version of Dijkstra’s algorithm, which
is motivated by the balls-and-strings model analogy of Figs. 2.11 and 2.12. In
particular, the algorithm may be interpreted as alternately lifting the model
upward from the origin (the following Step 1), and pulling the model downward
from the destination (the following Step 2). The algorithm maintains a price
vector p and two node subsets W1 and Wt . Initially, p satisfies the CS condition

pi ≤ aij + pj , ∀ (i, j) ∈ A, (2.20)

Sec. 2.8 Notes, Sources, and Exercises 107

W1 = {1}, and Wt = {t}. One may view W1 and Wt as the sets of permanently
labeled nodes from the origin and from the destination, respectively. The algo-
rithm terminates when W1 and Wt have a node in common. The typical iteration
is as follows:
Step 1 (Forward Step): Find

γ + = min{aij + pj − pi | (i, j) ∈ A, i ∈ W1 , j ∈
/ W1 }

and let
/ W1 | γ + = aij + pj − pi for some i ∈ W1 }.
V1 = {j ∈

Set 
pi + γ + if i ∈ W1 ,
pi :=
pi if i ∈
/ W1 .

Set
W1 := W1 ∪ V1 .

If W1 and Wt have a node in common, terminate the algorithm; otherwise, go to

Step 2.
Step 2 (Backward Step): Find

γ − = min{aji + pi − pj | (j, i) ∈ A, i ∈ Wt , j ∈
/ Wt }

and let
/ Wt | γ − = aji + pi − pj for some i ∈ Wt }.
Vt = {j ∈

Set 
pi − γ − if i ∈ Wt ,
pi :=
pi if i ∈
/ Wt .

Set
Wt := Wt ∪ Vt .

If W1 and Wt have a node in common, terminate the algorithm; otherwise, go to

Step 1.
(a) Show that throughout the algorithm, the condition (2.20) is maintained.
Furthermore, for all i ∈ W1 , p1 − pi is equal to the shortest distance from 1
to i. Similarly, for all i ∈ Wt , pi − pt is equal to the shortest distance from
i to t. Hint: Show that if i ∈ W1 , there exists a path from 1 to i such that
pm = amn + pn for all arcs (m, n) of the path.
(b) Show that the algorithm terminates and that upon termination, p1 − pt is
equal to the shortest distance from 1 to t.
(c) Show how the algorithm can be implemented so that its running time is
O(N 2 ). Hint: Let dmn denote the shortest distance from m to n. Maintain
the labels

vj+ = min{d1i + aij | i ∈ W1 , (i, j) ∈ A}, ∀j∈

/ W1 ,
108 The Shortest Path Problem Chap. 2

vj− = min{aji + dit | i ∈ Wt , (j, i) ∈ A}, ∀j∈

/ Wt .

Let p0j be the initial price of node j. Show that

 
   
γ + = min min vj+ + p0j , pt + min vj+ + djt −p1 , (2.21)
j ∈W
/ 1 , j ∈W
/ t j ∈W
/ 1 , j∈Wt

 
−
   
γ = min min vj− − p0j , −p1 + min vj− + d1j + pt .
j ∈W
/ 1 , j ∈W
/ t j∈W1 , j ∈W
/ t
(2.22)
Use these relations to calculate γ + and γ − in O(N ) time.
(d) Show how the algorithm can be implemented using binary heaps so that
its running time is O(A log N ). Hint: One possibility is to use four heaps
to implement the minimizations in Eqs. (2.21) and (2.22).
(e) Apply the two-sided version of Dijkstra’s algorithm with arc lengths aij +
pj − pi of Exercise 2.14, and with the termination criterion of part (c) of
that exercise. Show that the resulting algorithm is equivalent to the one of
the present exercise.

2.21

Consider the all-pairs shortest path problem, and suppose that the minimum
distances dij to go from any i to any j have been found. Suppose that a single
arc length amn is reduced to a value amn < amn . Show that if dnm + amn ≥ 0,
the new shortest distances can be obtained by

dij = min{dij , dim + amn + dnj }.

What happens if dnm + amn < 0?

2.22 (The Doubling Algorithm)

The doubling algorithm for solving the all-pairs shortest path problem is given by

aij if (i, j) ∈ A,
1
Dij = 0 if i = j,
∞ otherwise,


2k
k
minm Dim k
+ Dmj if i = j, k = 1, 2, . . . ,
log(N − 1),
Dij =
0 if i = j, k = 1, 2, . . . ,
log(N − 1).

Show that for i = j, Dij

k
gives the shortest distance from i to j using paths with
 
2 k−1
arcs or fewer. Show also that the running time is O N 3 log m∗ , where m∗
is the maximum number of arcs in a shortest path.
Sec. 2.8 Notes, Sources, and Exercises 109

2.23 (Dynamic Programming)

Consider the dynamic programming problem of Example 2.2. The standard dy-
namic programming algorithm is given by the recursion


Jk (xk ) = min gk (xk , uk ) + Jk+1 (xk+1 ) , k = 0, . . . , N − 1,
uk

starting with
JN (xN ) = G(xN ).

(a) In terms of the shortest path reformulation in Fig. 2.1, interpret Jk (xk ) as
the shortest distance from node xk at stage k to the terminal node t.
(b) Show that the dynamic programming algorithm can be viewed as a spe-
cial case of the generic label correcting algorithm with a special order for
selecting nodes to exit the candidate list.
(c) Assume that gk (xk , uk ) ≥ 0 for all xk , uk , and k. Suppose that by us-
ing some heuristic we can construct a “good” suboptimal control sequence
(u0 , u1 , . . . , uN −1 ). Discuss how to use this sequence for initialization of a
single origin/single destination label correcting algorithm (cf. the discussion
of Section 2.5).

2.24 (Forward Dynamic Programming)

Given a problem of finding a shortest path from node s to node t, we can obtain
an equivalent “reverse” shortest path problem, where we want to find a shortest
path from t to s in a graph derived from the original by reversing the direction of
all the arcs, while keeping their length unchanged. Apply this transformation to
the dynamic programming problem of Example 2.2 and Exercise 2.23, and derive
a dynamic programming algorithm that proceeds forwards rather than backwards
in time.

2.25 (k Shortest Node-Disjoint Paths)

The purpose of this exercise, due to Castañon [1990], is to formulate a class of

multiple shortest path problems and to indicate the method for their solution.
Consider a graph with an origin 1, a destination t, and a length for each arc.
We want to find k paths from 1 to t which share no node other 1 and t and
which are such that the sum of the k path lengths is minimum. Formulate this
problem as a minimum cost flow problem. (For an auction algorithm that solves
this problem, see Bertsekas and Castañon [1993c].) Hint: Replace each node i
other than 1 and t with two nodes i and i and a connecting arc (i, i ) with flow
bounds 0 ≤ xii ≤ 1.
110 The Shortest Path Problem Chap. 2

2.26 (k-Level Shortest Path Problems)

The purpose of this exercise, due to Shier [1979], and Guerriero, Lacagnina,
Musmanno, and Pecorella [1997], is to introduce an approach for extending the
generic algorithm to the solution of a class of multiple shortest path problems.
Consider the single origin/many destinations shortest path context, where node
1 is the origin, assuming that no cycles of negative length exist. Let di (1) denote
the shortest distance from node 1 to node i. Sequentially, for k = 2, 3, . . ., denote
by di (k) the minimum of the lengths of paths from 1 to i that have length greater
than di (k − 1) [if there is no path from 1 to i with length greater than di (k − 1),
then di (k) = ∞]. We call di (k) the k-level shortest distance from 1 to i.
(a) Show that for k > 1, {di (k) | i = 1, . . . , N } are the k-level shortest distances
if and only if di (k − 1) ≤ di (k) with strict inequality if di (k − 1) < ∞, and
furthermore


di (k) = min li (k, j) + aij , i = 1, . . . , N,
(i,j)∈A

where

di (k − 1) if dj (k − 1) < di (k − 1) + aij ,
li (k, j) =
di (k) if dj (k − 1) = di (k − 1) + aij .

(b) Extend the generic shortest path algorithm of Section 2.2 so that it simul-
taneously finds the k-level shortest distances for all k = 1, 2, . . . , K, where
K is some positive integer.

2.27 (Clustering)

We have a set of N objects 1, . . . , N arranged in a given order. We want to

group these objects in clusters that contain consecutive objects. For each subset
i, i + 1, . . . , i + k, there is an associated cost c(i, k). We want to find the grouping
that minimizes the sum of the clusters’ cost. Use the ideas of the paragraphing
problem (Example 2.4) to formulate this problem as a shortest path problem.

2.28 (Path Bottleneck Problem)

Consider the framework of the shortest path problem. For any path P , define
the bottleneck arc of P as an arc that has maximum length over all arcs of P .
Consider the problem of finding a path connecting two given nodes and having
minimum length of bottleneck arc. Derive an analog of Prop. 2.1 for this problem.
Consider also a single origin/all destinations version of this problem. Develop an
analog of the generic algorithm of Section 2.2 and prove an analog of Prop. 2.2.
Hint: Replace di + aij with max{di , aij }.
Sec. 2.8 Notes, Sources, and Exercises 111

2.29 (Shortest Path Problems with Negative Cycles)

Consider the problem of finding a simple forward path between an origin and a
destination node that has minimum length. Show that even if there are negative
cycles, the problem can be formulated as a minimum cost flow problem involving
node throughput constraints of the form


0≤ xij ≤ 1, ∀ i.
{j|(i,j)∈A}

2.30 (Minimum Weight Spanning Trees)

Given a graph (N , A) and a weight wij for each arc (i, j), consider the problem
of finding a spanning tree with minimum sum of arc weights. This is not a
shortest path problem and in fact it is not even a special case of the minimum
cost flow problem. However, it has a similar graph structure to the one of the
shortest path problem. Note that the orientation of the arcs does not matter
here. In particular, if (i, j) and (j, i) are arcs, any one of them can participate
in a spanning tree solution, and the arc having greater weight can be a priori
eliminated.
(a) Consider the problem of finding a shortest path from node 1 to all nodes
with arc lengths equal to wij . Give an example where the shortest path
spanning tree is not a minimum weight spanning tree.
(b) Let us define a fragment to be a subgraph of a minimum weight spanning
tree; for example the subgraph consisting of any subset of nodes and no
arcs is a fragment. Given a fragment F , let us denote by A(F ) the set of
arcs (i, j) such that either i or j belong to F , and if (i, j) is added to F
no cycle is closed. Show that if F is a fragment, then by adding to F an
arc of A(F ) that has minimum weight over all arcs of A(F ) we obtain a
fragment.
(c) Consider a greedy algorithm that starts with some fragment, and at each
iteration, adds to the current fragment F an arc of A(F ) that has minimum
weight over all arcs of A(F ). Show that the algorithm terminates with a
minimum weight spanning tree.
(d) Show that the complexity of the greedy algorithm is O(N A), where N is
the number of nodes and A is the number of arcs.
(e) The Prim-Dijkstra algorithm is the special case of the greedy algorithm
where the initial fragment consists of a single node. Provide an O(N 2 ), im-
plementation of this algorithm. Hint: Together with the kth fragment Fk ,
maintain for each j ∈/ Fk the node nk (i) ∈ Fk such that the arc connecting
j and nk (i) has minimum weight.
112 The Shortest Path Problem Chap. 2

2.31 (Shortest Path Problems with Losses)

Consider a vehicle routing/shortest path-like problem where a vehicle wants to

go on a forward path from an origin node 1 to a destination node t in a graph
that has no forward cycles. For each arc (i, j) there is a given length aij , but
there is also a given probability pij ∈ [0, 1] that the vehicle will be destroyed in
crossing the arc. The length of a path is now a random variable, and is equal
to the sum of the arc lengths on the path up to the time the vehicle reaches its
destination or gets destroyed, whichever comes first. We want to find a forward
path P = (1, i1 , . . . , ik , t) whose expected length, given by
  
p1i1 a1i1 + pi1 i2 ai1 i2 + pi2 i3 (· · · + pik t aik t ) · · · ,

is minimized, where pij = 1 − pij is the probability of survival in crossing the

arc (i, j). Give an algorithm of the dynamic programming type for solving this
problem (cf. Exercise 2.5). Does the problem always make sense when the graph
has some forward cycles?

2.32

Consider the one origin-all destinations problem and the generic algorithm of
Section 2.2. Assume that there exists a path that starts at node 1 and contains
a cycle with negative length. Assume also that the generic algorithm is operated
so that if a given node belongs to the candidate list for an infinite number of
iterations, then it also exits the list an infinite number of times. Show that there
exists at least one node j such that the sequence of labels dj generated by the
algorithm diverge to −∞. Hint: Argue that if the limits dj of all the label nodes
are finite, then we have dj ≤ di + aij for all arcs (i, j).

2.33 (A Modified Auction Algorithm for Shortest Paths)

Consider the problem of finding a shortest path from node 1 to a node t, as-
suming that there exists at least one such path and that all cycles have positive
length. This exercise deals with a modified version of the auction algorithm,
which was developed in Bertsekas [1992b], motivated by a similar earlier algo-
rithm by Cerulli, De Leone, and Piacente [1994]. This modified version aims to
use larger price increases than the original method. The algorithm maintains a
price vector p and a simple path P that starts at the origin, and is initialized
with P = (1) and any price vector p satisfying

p1 = ∞,

pi ≤ aij + pj , ∀ (i, j) ∈ A with i = 1.

The algorithm terminates when the destination t becomes the terminal node of
P . To describe the algorithm, define

A(i) = {j | (i, j) ∈ A} ∪ {i}, ∀ i ∈ N,

Sec. 2.8 Notes, Sources, and Exercises 113

aii = 0, ∀ i ∈ N.

The typical iteration is as follows:

Let i be the terminal node of P , and let ji be such that


ji = arg min aij + pj ,
j∈A(i)

with the extra requirement that ji = i whenever possible; that is, we choose
ji = i whenever the minimum above is attained for some j = i. Set


pji := min aij + pj − aiji .
j∈A(i), j=ji

If ji = i contract P ; otherwise extend P by node ji .

Note that if a contraction occurs, we have ji = i = 1 and the price of the terminal
node pi is strictly increased. Note also that when an extension occurs from the
terminal node i to a neighbor ji = i, the price pji may be increased strictly, while
in the original auction algorithm there is no price change. Furthermore, the CS
condition pi ≤ aij + pj for all (i, j) is not maintained. Show that:
(a) The algorithm maintains the conditions

πi = aij + pj , ∀ (i, j) ∈ P,

πi = pi , ∀i∈
/ P,

where  
πi = min pi , min {aij + pj } , ∀ i ∈ N.
{j|(i,j)∈A}

(b) Throughout the algorithm, P is a shortest path between its endnodes. Hint:
Show that if P̃ is another path with the same endnodes, we have

 
Length of P̃ − Length of P = (πk − pk ) − (πk − pk )
{k|k∈P̃ , k∈P
/ } {k|k∈P, k∈
/ P̃ }

≥ 0.

(c) The algorithm terminates with a shortest path from 1 to t. Note: This is
challenging. A proof is given in Bertsekas [1992b].
(d) Convert the shortest path problem to an equivalent assignment problem
for which the conditions of part (a) are the complementary slackness con-
ditions. Show that the algorithm is essentially equivalent to a naive auction
algorithm applied to the equivalent assignment problem.
114 The Shortest Path Problem Chap. 2

2.34 (Continuous Space Shortest Path Problems)

 
Consider a continuous-time dynamic system whose state x(t) = x1 (t), x2 (t)
evolves in two-dimensional space according to the differential equations

ẋ1 (t) = u1 (t), ẋ2 (t) = u2 (t)

 
where for each time t, u(t) = u1 (t), u2 (t) is a two-dimensional control vector
with unit norm. We want to find a state trajectory that starts at a given point
x(0), ends at another given point x(T ), and minimizes
 T
 
r x(t) dt,
0

where r(·) is a given nonnegative and continuous function. The final time T and
the control trajectory {u(t) | 0 ≤ t ≤ T } are subject to optimization. Suppose
we discretize the plane with a mesh of size δ that passes through x(0) and x(T ),
and we introduce a shortest path problem of going from x(0) to x(T ) using moves
of the following type: from each mesh point x = (x1 , x2 ) we can go to each of
the mesh points (x1 + δ, x2 ), (x1 − δ, x2 ), (x1 , x2 + δ), and (x1 , x2 − δ), at a cost
r(x)δ. Show by example that this is a bad discretization of the original problem
in the sense that the shortest distance need not approach the optimal cost of
the original problem as δ → 0. Note: This exercise illustrates a common pitfall.
The difficulty is that the control constraint set (the surface of the unit sphere)
should be finely discretized as well. For a proper treatment of the problem
of discretization, see the original papers by Gonzalez and Rofman [1985], and
Falcone [1987], the survey paper by Kushner [1990], the monograph by Kushner
and Dupuis [1992], and the references cited there. For analogs of the label setting
and label correcting algorithms of the present chapter, see the papers by Tsitsiklis
[1995], and by Polymenakos, Bertsekas, and Tsitsiklis [1998].
3

The Max-Flow Problem

Contents

3.1. The Max-Flow and Min-Cut Problems

3.1.1. Cuts in a Graph
3.1.2. The Max-Flow/Min-Cut Theorem
3.1.3. The Maximal and Minimal Saturated Cuts
3.1.4. Decomposition of Infeasible Network Problems
3.2. The Ford-Fulkerson Algorithm
3.3. Price-Based Augmenting Path Algorithms
3.3.1. A Price-Based Path Construction Algorithm
3.3.2. A Price-Based Max-Flow Algorithm
3.4. Notes, Sources, and Exercises

115
 116 The Max-Flow Problem Chap. 3

In this chapter, we focus on the max-flow problem introduced in Example

1.3 of Section 1.2. We have a graph (N , A) with flow bounds xij ∈ [bij , cij ]
for each arc (i, j), and two special nodes s and t. We want to maximize
the divergence out of s over all capacity-feasible flow vectors having zero
divergence for all nodes except s and t.
The max-flow problem arises in a variety of practical contexts and
also as a subproblem in the context of algorithms that solve other more
complex problems. For example, it can be shown that checking the exis-
tence of a feasible solution of a minimum cost flow problem, and finding a
feasible solution if one exists, is essentially equivalent to a max-flow prob-
lem (see Fig. 3.1, and Exercises 3.3 and 3.4). Furthermore, a number of
interesting combinatorial problems can be posed as max-flow problems (see
for example Exercises 3.8-3.10).
Like the shortest path problem, the max-flow problem embodies a
number of methodological ideas that are central to the more general min-
imum cost flow problem. In fact, whereas the shortest path problem can
be viewed as a minimum cost flow problem where arc capacities play no
role, the max-flow problem can be viewed as a minimum cost flow problem
where arc costs play no role. In this sense, the structures of the short-
est path and max-flow problems are complementary, and together provide
the foundation upon which much of the algorithmic methodology of the
minimum cost flow problem is built.
Central to the max-flow problem is the max-flow/min-cut theorem,
which is one of the most celebrated theorems of network optimization. In
Section 3.1, we derive this result, and we discuss some of its applications.
Later, in Chapter 4, we will interpret this result as a duality theorem
(see Exercise 4.4). In Section 3.2, we introduce a central algorithm for
solving the max-flow problem, the Ford-Fulkerson method. This is a fairly
simple method, which however can behave in interesting and surprising
ways. Much research has been devoted to developing clever and efficient
implementations of the Ford-Fulkerson method. We describe some of these
implementations in Sections 3.2 and 3.3, and in the exercises.

3.1 THE MAX-FLOW AND MIN-CUT PROBLEMS

The key idea in the max-flow problem is very simple: a feasible flow x can
be improved if we can find a path from s to t that is unblocked with respect
to x. Pushing a positive increment of flow along such a path results in larger
divergence out of s, while maintaining flow feasibility. Most (though not
all) of the available max-flow algorithms are based on iterative application
of this idea.
We may also ask the reverse question. If we can’t find an unblocked
path from s to t, is the current flow maximal? The answer is positive,
Sec. 3.1 The Max-Flow and Min-Cut Problems 117

Source Node 1 4 Sink Node

s1 > 0 s4 < 0
s3 = 0

Source Node 2 5 Sink Node

s2 > 0 s5 < 0

1 4
[0, s1 ] [0,- s4 ]

Source s 3 t Sink

[0, s2 ] [0,- s5 ]
2 5

Figure 3.1: Essential equivalence of the problem of finding a feasible solution of a

minimum cost flow problem and a max-flow problem. Given a set of divergences
si satisfying s = 0, and capacity intervals [0, cij ], consider the feasibility
i i
problem of finding a flow vector x satisfying
 
xij − xji = si , ∀ i ∈ N, (3.1)
{j|(i,j)∈A} {j|(j,i)∈A}

0 ≤ xij ≤ cij , ∀ (i, j) ∈ A. (3.2)

Denote by I + = {i | si > 0} the set of source nodes ({1, 2} in the figure) and by
I − = {i | si < 0} the set of sink nodes ({4, 5} in the figure). If both these sets
are empty, the zero vector is a feasible flow, and we are done. Otherwise, these
sets are both nonempty (since s = 0). We introduce a node s, and for all
i i
i ∈ I + , the arcs (s, i) with flow range [0, si ]. We also introduce a node t, and
for all i ∈ I − , the arcs (i, t) with flow range [0, −si ]. Now consider the max-flow
problem of maximizing the divergence out of s and into t, while observing the
capacity constraints. Then there exists a solution to the feasibility problem of

Eqs. (3.1) and (3.2), if and only if the maximum divergence out of s is equal to
i∈I + i
s . If this condition is satisfied, solutions of the feasibility problem are in
one-to-one correspondence with optimal solutions of the max-flow problem.
If the capacity constraints involve lower bounds, bij ≤ xij ≤ cij , we may
convert first the feasibility problem to one with zero lower flow bounds by a
translation of variables, which replaces each variable xij with a variable zij =
xij − bij .
Also, a max-flow problem can (in principle) be solved by an algorithm
that solves the feasibility problem (we try to find a sequence of feasible flows
with monotonically increasing divergence out of s, stopping with a maximum flow
when no further improvement is possible). In fact, this is the main idea of the
Ford-Fulkerson method, to be discussed in Section 3.2.
118 The Max-Flow Problem Chap. 3

although the reason is not entirely obvious. For a brief justification, con-
sider the minimum cost flow formulation of the max-flow problem, given in
Example 1.3, which involves the artificial feedback arc (t, s) (see Fig. 3.2).
Then, a cycle has negative cost if and only if it includes the arc (t, s), since
this arc has cost -1 and is the only arc with nonzero cost. By Prop. 1.2, if
a feasible flow vector x is not optimal, there must exist a simple cycle with
negative cost that is unblocked with respect to x; this cycle must consist
of the arc (t, s) and a path from s to t, which is unblocked with respect to
x. Thus, if there is no path from s to t that is unblocked with respect to a
given flow vector x, then there is no cycle of negative cost and x must be
optimal.

All cost coefficients are

zero except for ats

Source s t Sink

Artificial feedback arc

Cost coefficient = -1

Figure 3.2: Minimum cost flow formulation of a max-flow problem, involving

a feedback (t, s) arc with cost -1 and unconstrained arc flow (−∞ < xts < ∞).
For a nonoptimal flow x, there must exist a cycle that is unblocked with respect
to x and has negative cost. Since all arcs other than the feedback arc have zero
length, this cycle must contain the feedback arc. This implies that there must
exist a path from s to t, which is unblocked with respect to x. Many max-flow
algorithms push flow along such a path to iteratively improve an existing flow
vector x.

The max-flow/min-cut theorem and the Ford-Fulkerson algorithm,

to be described shortly, are based on the preceding ideas. However, rather
than appealing to Prop. 1.2 (whose proof relies on the notion of a conformal
decomposition), we couch the analysis of this chapter on first principles,
taking advantage of the simplicity of the max-flow problem. This will also
serve to develop some concepts that will be useful later. We first introduce
some definitions.
Sec. 3.1 The Max-Flow and Min-Cut Problems 119

3.1.1 Cuts in a Graph

A cut Q in a graph (N , A) is a partition of the node set N into two

nonempty subsets, a set S and its complement N − S. We use the notation

Q = [S, N − S].

Note that the partition is ordered in the sense that the cut [S, N − S] is
distinct from the cut [N − S, S]. For a cut Q = [S, N − S], we use the
notation


Q+ = (i, j) ∈ A | i ∈ S, j ∈
/S ,


Q− = (i, j) ∈ A | i ∈
/ S, j ∈ S ,

and we say that Q+ and Q− are the sets of forward and backward arcs of
the cut, respectively. We say that the cut Q is nonempty if Q+ ∪ Q− = Ø;
otherwise we say that Q is empty. We say that the cut [S, N − S] separates
node s from node t if s ∈ S and t ∈/ S. These definitions are illustrated in
Fig. 3.3.

Figure 3.3: Illustration of a cut

2 4
Q = [S, N − S],

1 6 where S = {1, 2, 3}. We have

Q+ = {(2, 4), (1, 6)},

3 5

Q− = {(4, 1), (6, 3), (5, 3)}.

Given a flow vector x, the flux across a nonempty cut Q = [S, N − S]

is defined to be the total net flow coming out of S, i.e., the scalar
 
F (Q) = xij − xij .
(i,j)∈Q+ (i,j)∈Q−

Let us recall from Section 1.1.2 the definition of the divergence of a node i:
 
yi = xij − xji , ∀ i ∈ N.
{j|(i,j)∈A} {j|(j,i)∈A}
120 The Max-Flow Problem Chap. 3

The following calculation shows that F (Q) is also equal to the sum of the
divergences yi of the nodes in S:
 
F (Q) = xij − xij
{(i,j)∈A|i∈S,j ∈S}
/ {(i,j)∈A|i∈S,j∈S}
/
⎛ ⎞
  
= ⎝ xij − xji ⎠ (3.3)
i∈S {j|(i,j)∈A} {j|(j,i)∈A}

= yi .
i∈S

(The second equality holds because the flow of an arc with both end nodes
in S cancels out within the parentheses; it appears twice, once with a
positive and once with a negative sign.)
Given lower and upper flow bounds bij and cij for each arc (i, j), the
capacity of a nonempty cut Q is
 
C(Q) = cij − bij . (3.4)
(i,j)∈Q+ (i,j)∈Q−

Clearly, for any capacity-feasible flow vector x, the flux F (Q) across Q is
no larger than the cut capacity C(Q). If F (Q) = C(Q), then Q is said to
be a saturated cut with respect to x; the flow of each forward (backward)
arc of such a cut must be at its upper (lower) bound. By convention, every
empty cut is also said to be saturated. The following is a simple but useful
result.

Proposition 3.1: Let x be a capacity-feasible flow vector, and let s

and t be two nodes. Then exactly one of the following two alternatives
holds:
(1) There exists a simple path from s to t that is unblocked with
respect to x.
(2) There exists a saturated cut that separates s from t.

Proof: The proof is obtained by constructing an algorithm that terminates

with either a path as in (1) or a cut as in (2). This algorithm is a special
case of a general method, known as breadth-first search, and used to find a
simple path between two nodes in a graph (see Exercise 3.2). The algorithm
generates a sequence of node sets {Tk }, starting with T0 = {s}; each set Tk
represents the set of nodes that can be reached from s with an unblocked
path of k arcs.
Sec. 3.1 The Max-Flow and Min-Cut Problems 121

Unblocked Path Search Algorithm

For k = 0, 1, . . ., given Tk , terminate if either Tk is empty or t ∈ Tk ;

otherwise, set


Tk+1 = n ∈
/ ∪ki=0 Ti | there is a node m ∈ Tk , and either an arc (m, n)

with xmn < cmn , or an arc (n, m) with bnm < xnm

and mark each node n ∈ Tk+1 with the label “(m, n)” or “(n, m),”
where m is a node of Tk and (m, n) or (n, m) is an arc with the property
stated in the above equation, respectively.

Figure 3.4 illustrates the preceding algorithm. Since the algorithm

terminates if Tk is empty, and Tk must consist of nodes not previously
included in ∪k−1
i=0 Ti , the algorithm must eventually terminate. Let S be the
union of the sets Ti upon termination. There are two possibilities:
(a) The final set Tk contains t, in which case, by tracing labels backward
from t, a simple unblocked path P from s to t can be constructed.
The forward arcs of P are of the form (m, n) with xmn < cmn and
the label of n being “(m, n)”; the backward arcs of P are of the
form (n, m) with bnm < xnm and the label of n being “(n, m).” Any
cut separating s from t must contain a forward arc (m, n) of P with
xmn < cmn or a backward arc (n, m) of P with bnm < xnm , and
therefore cannot be saturated. Thus, the result is proved in this case.
(b) The final set Tk is empty, in which case from the equation defining Tk ,
it can be seen that the cut Q = [S, N − S] is saturated and separates
s from t. To show that there is no simple unblocked path from s
to t, note that any such path must have either an arc (m, n) ∈ Q+
with xmn < cmn or an arc (n, m) ∈ Q− with bnm < xnm , which is
impossible, since Q is saturated.
Q.E.D.

Exercise 3.11 provides some variations of Prop. 3.1. In particular, in

place of s and t, one may use two disjoint subsets of nodes N + and N − .
Furthermore, “simple path” in alternative (1) may be replaced by “path.”

3.1.2 The Max-Flow/Min-Cut Theorem

Consider now the max-flow problem, where we want to maximize the diver-
gence out of s over all capacity-feasible flow vectors having zero divergence
for all nodes other than s and t. Given any such flow vector and any cut
Q separating s from t, the divergence out of s is equal to the flux across Q
[cf. Eq. (3.3)], which in turn is no larger than the capacity of Q. Thus, if
122 The Max-Flow Problem Chap. 3

(lower bound, flow, upper bound)

shown next to each arc

(1,1,2) T2 Figure 3.4: Illustration of the un-

2 4
blocked path search algorithm for
(-1,0,1) T1 (0,0,1) finding an unblocked path from node
(0,0,1)
1 to node 6, or a saturated cut sep-
T0 1 (0,1,1) arating 1 from 6. The triplet (lower
5
(1,2,2) bound, flow, upper bound) is shown
(0,0,1)
next to each arc. The figure shows
(0,1,1)
the successive sets Tk generated by
3 6
(1,1,2) the algorithm. In case (a) there ex-
T3 ists a unblocked path from 1 to 6,
(a) namely the path (1, 3, 5, 6). In case
(b), where the flow of arc (6, 5) is
at the lower bound rather than the
upper bound, there is the saturated
(1,1,2) T2 cut [S, N − S] separating 1 from 6,
2 4 where S = {1, 2, 3, 4, 5} is the union
(-1,0,1) T1 (0,0,1) of the sets Tk . Note that the algo-
(0,0,1)
rithm works for any arc flows, and,
T0 1 (0,1,1) in particular, does not require that
5
the nodes other than the start node
(1,2,2) 1 and the end node 6 have zero di-
(0,0,1)
(0,0,1) vergence.
3 6
(1,1,2)

(b)

the max-flow problem is feasible, we have

Maximum Flow ≤ Capacity of Q. (3.5)

The following max-flow/min-cut theorem asserts that equality is attained

for some Q. Part (a) of the theorem assumes the existence of an optimal
solution to the max-flow problem. This assumption need not be satisfied;
indeed it is possible that the max-flow problem has no feasible solution at
all (consider a graph consisting of a single two-arc path from s to t, the
arcs of which have disjoint feasible flow ranges). In Chapter 5, however,
we will show using the theory of the simplex method (see Prop. 5.7), that
the max-flow problem (and indeed every minimum cost flow problem) has
an optimal solution if it has at least one feasible solution. [Alternatively,
this can be shown using a fundamental result of mathematical analysis,
the Weierstrass theorem, which states that a continuous function attains
a maximum over a nonempty and compact set (see Appendix A and the
sources given there).] If the lower flow bound is zero for every arc, the max-
flow problem has at least one feasible solution, namely the zero flow vector.
Sec. 3.1 The Max-Flow and Min-Cut Problems 123

Thus the theory of Chapter 5 (or the Weierstrass theorem) guarantees that
the max-flow problem has an optimal solution in this case. This is stated
as part (b) of the following theorem, even though its complete proof must
await the developments of Chapter 5.

Proposition 3.2: (Max-Flow/Min-Cut Theorem)

(a) If x∗ is an optimal solution of the max-flow problem, then the
divergence out of s corresponding to x∗ is equal to the minimum
cut capacity over all cuts separating s from t.
(b) If all lower arc flow bounds are zero, the max-flow problem has
an optimal solution, and the maximal divergence out of s is equal
to the minimum cut capacity over all cuts separating s from t.

Proof: (a) Let F ∗ be the value of the maximum flow, that is, the diver-
gence out of s corresponding to x∗ . There cannot exist an unblocked path
P from s to t with respect to x∗ , since by increasing the flow of the forward
arcs of P and by decreasing the flow of the backward arcs of P by a com-
mon positive increment, we would obtain a flow vector with a divergence
out of s larger than F ∗ . Therefore, by Prop. 3.1, there must exist a cut
Q, that is saturated with respect to x∗ and separates s from t. The flux
across Q is equal to F ∗ and is also equal to the capacity of Q [since Q is
saturated; see Eqs. (3.3) and (3.4)]. Since we know that F ∗ is less or equal
to the minimum cut capacity [cf. Eq. (3.5)], the result follows.
(b) See the discussion preceding the proposition. Q.E.D.

3.1.3 The Maximal and Minimal Saturated Cuts

Given an optimal solution x∗ of the max-flow problem, there may exist

several saturated cuts [S, N − S] separating s and t. We will show that out
of these cuts, there exists one, called maximal , corresponding to the union
of the sets S. Similarly, there is a minimal saturated cut, corresponding to
the intersection of the sets S. (The maximal and minimal cuts coincide if
and only if there is a unique saturated cut.)
Indeed, let S be the union of all node sets S such that [S, N − S] is
a saturated cut separating s and t. Consider the cut

Q = [S, N − S].
+
Clearly Q separates s and t. If (i, j) ∈ Q , then we have x∗ij = cij because
i belongs to one of the sets S such that [S, N − S] is a saturated cut, and j
+
does not belong to S since j ∈/ S. Thus we have x∗ij = cij for all (i, j) ∈ Q .
124 The Max-Flow Problem Chap. 3

−
Similarly, we obtain x∗ij = bij for all (i, j) ∈ Q . Thus Q is a saturated cut
separating s and t, and in view of its definition, it is the maximal such cut.
By using set intersection in place of set union in the preceding argument, it
is seen that we can similarly form the minimal saturated cut that separates
s and t.
The maximal and minimal saturated cuts can be used to deal with
infeasibility in the context of various network flow problems, as we discuss
next.

3.1.4 Decomposition of Infeasible Network Problems

Consider the minimization of a separable cost function of the flow vector

x, 
fij (xij ),
(i,j)∈A

subject to conservation of flow constraints

 
xij − xji = si , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

and capacity constraints

0 ≤ xij ≤ cij , ∀ (i, j) ∈ A.

We assume that the scalars si are given and satisfy i∈N si = 0, but that
the problem is infeasible, because the capacities cij are not sufficiently large
to carry all the supply from the set of supply nodes

I + = {i | si > 0}

to the set of demand nodes

I − = {i | si < 0}.

Then it may make sense to minimize the cost function over the set of all
maximally feasible flows, which is the set of flow vectors x whose diver-
gences  
yi = xij − xji
{j|(i,j)∈A} {j|(j,i)∈A}

satisfy
yi ≥ 0 if i ∈ I + ,
yi ≤ 0 if i ∈ I − ,
yi = 0 if i ∈
/ I + ∪ I −,
 Sec. 3.2 The Ford-Fulkerson Algorithm 125

and minimize 
|si − yi |.
i∈N

Thus, roughly, a flow vector is maximally feasible if it is capacity-feasible,

and it satisfies as much of the given demand as possible by using as much
of the given supply as possible.
Note that we can find a maximally feasible flow x∗ by solving the
max-flow problem given in Fig. 3.1. The vector x∗ defines corresponding
minimal and maximal saturated cuts

[Smin , N − Smin ], [Smax , N − Smax ],

respectively, separating the supply node set P from the demand node set D.
Furthermore, the flows of all arcs (i, j) that belong to these cuts are equal to
x∗ij for every maximally feasible flow vector. It can now be seen that given
x∗ , we can decompose the problem of minimizing the cost function over the
set of maximally feasible flows into two or three feasible and independent
subproblems, depending on whether Smin = Smax or not. The node sets of
these problems are Smin , N − Smax , and Smax − Smin , (if Smax = Smin ).
The supplies for these problems are appropriately adjusted to take into
account the arc flows x∗ij for the arcs (i, j) of the corresponding cuts, as
illustrated in Fig. 3.5.

3.2 THE FORD-FULKERSON ALGORITHM

In this section, we focus on a fundamental algorithm for solving the max-

flow problem. This algorithm is of the primal cost improvement type,
because it improves the primal cost (the divergence out of s) at every
iteration. The idea is that, given a feasible flow vector x (i.e., one that is
capacity-feasible and has zero divergence out of every node other than s
and t), and a path P from s to t, which is unblocked with respect to x, we
can increase the flow of all forward arcs (i, j) of P and decrease the flow of
all backward arcs (i, j) of P . The maximum increment of flow change is


δ = min {cij − xij | (i, j) ∈ P + }, {xij − bij | (i, j) ∈ P − } ,

where P + is the set of forward arcs of P and P − is the set of backward

arcs of P . The resulting flow vector x, given by
⎧
⎨ xij + δ if (i, j) ∈ P + ,
xij = xij − δ if (i, j) ∈ P − ,
⎩
xij otherwise,
126 The Max-Flow Problem Chap. 3

Max-Flow/Capacity
shown next to each arc.
All lower flow bounds are 0.
1/1
1 5
2/3 1/1 1/1 2/3

6 0/1 2/3 0/1 6

s 3 4 t

2/3 1/1 1/1 2/3

2 6
1/1

Minimal Maximal
Saturated Cut Saturated Cut

3.5: Decomposition of the problem of minimizing a separable cost func-

Figure
tion f (x ) over the set of maximally feasible flow vectors into three
(i,j)∈A ij ij
(feasible) optimization problems. The problem here is to send 6 units of flow from
node s to node t, while satisfying capacity constraints [0, cij ] and minimizing a
cost function f (x ). In this example, all arcs have capacity 1, ex-
(i,j)∈A ij ij
cept for arc (3, 4) and the incident arcs to nodes s and t, which have capacity 3.
The problem is infeasible, so we consider optimization over all maximally feasible
solutions. We solve the max-flow problem from s to t, and we obtain the corre-
sponding minimal and maximal saturated cuts, as shown in the figure. Note that
the flows of the arcs across these cuts are unique, although the max-flow vector
is not unique.
We can now decompose the original (infeasible) optimization problem into
three (feasible) optimization problems, each with the cost function f (x ),
(i,j) ij ij
where the summation is over the relevant set of arcs. These problems are:
(1) The problem involving the nodes s, 1, and 2, with conservation of flow
constraints

xs1 + xs2 = 4, −x21 − xs1 = −2, x21 − xs2 = −2.

(2) The problem involving the nodes 3 and 4, with conservation of flow con-
straint (for both nodes) x34 = 2.
(3) The problem involving the nodes 5, 6, and t, with conservation of flow
constraints

x5t + x56 = 2, x6t − x56 = 2, −x5t − x6t = −4.

Note that while in this example the 2nd problem is trivial (has only one feasible
solution), the 1st and 3rd problems have multiple feasible solutions.
Sec. 3.2 The Ford-Fulkerson Algorithm 127

is feasible, and it has a divergence out of s that is larger by δ than the

divergence out of s corresponding to x. We refer to P as an augmenting
path, and we refer to the operation of replacing x by x as a flow augmenta-
tion along P . Such an operation may also be viewed as a modification of x
along the negative cost cycle consisting of P and an artificial arc (t, s) that
has cost −1; see the formulation of the max-flow problem as a minimum
cost flow problem in Fig. 3.2, and the discussion at the beginning of Section
3.1.
The Ford-Fulkerson algorithm starts with a feasible flow vector. If
the lower flow bound is zero for all arcs, the zero flow vector can be used
as a starting vector; otherwise, a preliminary phase is needed to obtain a
feasible starting flow vector. This involves solving an auxiliary max-flow
problem with zero lower flow bounds starting from the zero flow vector
and using the Ford-Fulkerson algorithm described below (cf. Fig. 3.1 and
Exercise 3.4). At each iteration the algorithm has a feasible flow vector
and uses the unblocked path search method, given in the proof of Prop.
3.1, to either generate a new feasible flow vector with larger divergence out
of s or terminate with a maximum flow and a minimum capacity cut.

Iteration of Ford-Fulkerson Algorithm

Use the unblocked path search method to either

(1) find a saturated cut separating s from t or
(2) find an unblocked path P with respect to x starting from s and
ending at t.
In case (1), terminate the algorithm; the current flow vector solves the
max-flow problem. In case (2), perform an augmentation along P and
go to the next iteration.

Figure 3.6 illustrates the Ford-Fulkerson algorithm. Based on the pre-

ceding discussion, we see that with each augmentation, the Ford-Fulkerson
algorithm improves the primal cost (the divergence out of s) by the augmen-
tation increment δ. Thus, if δ is bounded below by some positive number,
the algorithm can execute only a finite number of iterations and must ter-
minate with an optimal solution. In particular, if the arc flow bounds are
integer and the initial flow vector is also integer, δ is a positive integer at
each iteration, and the algorithm terminates. The same is true even if the
arc flow bounds and the initial flow vector are rational; by multiplication
with a suitably large integer, one can scale these numbers up to integer
while leaving the problem essentially unaffected.
On the other hand, if the problem data are irrational, proving termi-
nation of the Ford-Fulkerson algorithm is nontrivial. The proof (outlined
in Exercise 3.12) depends on the use of the specific unblocked path search
128 The Max-Flow Problem Chap. 3

2 2
[0,4] [0,1] 1 1
[0,1]
1 4 5 1 5
[0,2] [0,1] [0,5]
[0,2] [0,3]
3

2 2
1 1 1 1
1 0 1 1
0 0 4 5 4 5
0
0 0
3

2
2 1
1 2
1 0 0 4 5 1 4 5
1
0 0 2 2
3 3

2 2
2 1 1
1
1 0 0 4 5 1 1 4 5
3 1
2 2
3 1
3

2
3 1

1 1 0 1 4 5
4
2 3
3

Figure 3.6: Illustration of the Ford-Fulkerson algorithm for finding a maximum

flow from node s = 1 to node t = 5. The arc flow bounds are shown next to the arcs
in the top left figure, and the starting flow is zero. The sequence of successive flow
vectors is shown on the left, and the corresponding sequence of augmentations is
shown on the right. The saturated cut obtained is [{1, 2, 3}, {4, 5}]. The capacity
of this cut as well as the maximum flow is 5.

method of Prop. 3.1; this method (also referred to as breadth-first search,

see Exercise 3.2) yields augmenting paths with as few arcs as possible (see
Exercises 3.2 and 3.12). If unblocked paths are constructed using a dif-
ferent method, then, surprisingly, the Ford-Fulkerson algorithm need not
terminate, and the generated sequence of divergences out of s may con-
verge to a value strictly smaller than the maximum flow (for an example,
see Exercise 3.7, and for a different example, see Ford and Fulkerson [1962],
or Papadimitriou and Steiglitz [1982], p. 126, or Rockafellar [1984], p. 92).
Even with integer problem data, if the augmenting paths are constructed
Sec. 3.2 The Ford-Fulkerson Algorithm 129

using a different unblocked path search method, the Ford-Fulkerson algo-

rithm may require a very large (pseudopolynomial) number of iterations to
terminate; see Fig. 3.7.

[0,C] 2 [0,C]

1 [0,1] 4
Figure 3.7: An example showing
[0,C] [0,C] that if the augmenting paths used
3 in the Ford-Fulkerson algorithm do
not have a number of arcs that
is as small as possible, the num-
ber of iterations may be very large.
2 Augmenting Path for Odd
1 Here, C is a large integer. The
Numbered Iterations
maximum flow is 2C, and can be
1 1 4 produced after a sequence of 2C
augmentations using the three-arc
augmenting paths shown in the fig-
1
3 ure. Thus, the running time is
pseudopolynomial (it is proportio-
nal to C).
If on the other hand the two-
Augmenting Path for Even
2 Numbered Iterations arc augmenting paths (1, 2, 4) and
(1, 3, 4) are used, only two augmen-
-1 1 tations are needed.
1 4

1
3

Polynomial Max-Flow Algorithms

Using “shortest” augmenting paths (paths with as few arcs as possible) not
only guarantees termination of the Ford-Fulkerson algorithm. It turns out
that it also results in polynomial running time, as the example of Fig. 3.7
illustrates. In particular, the number of augmentations of the algorithm
with shortest augmenting paths can be estimated as O(N A); see Exercise
3.12. This yields an O(N A2 ) running time to solve the problem, since
each augmentation requires O(A) operations to execute the unblocked path
search method and to carry out the subsequent flow update.
Much research has been devoted to developing max-flow algorithms
with better than O(N A2 ) running time. The algorithms that we will discuss
can be grouped into two main categories:
(a) Variants of the Ford-Fulkerson algorithm, which use special data
structures and preprocessing calculations to generate augmenting paths
efficiently. We will describe some algorithms of this type in what fol-
lows in this chapter.
130 The Max-Flow Problem Chap. 3

(b) Algorithms that depart from the augmenting path approach, but in-
stead move flow from the source to the sink in a less structured fash-
ion than the Ford-Fulkerson algorithm. These algorithms, known as
preflow-push methods, will be discussed in Section 7.3. Their underly-
ing mechanism is related to the one of the auction algorithm described
in Section 1.3.3.
The algorithms that have the best running times at present are the preflow-
push methods. In particular, in Section 7.3 we will demonstrate an O(N 3 )
running time for one of these methods, and we will describe another method
with an O(N 2 A1/2 ) running time. Preflow-push algorithms with even bet-
ter running times exist (see the discussion in Chapter 7). It is unclear,
however, whether the best preflow-push methods outperform in practice
the best of the Ford-Fulkerson-like algorithms of this chapter.
In the remainder of this chapter, we will discuss efficient variants of
the Ford-Fulkerson algorithm. These variants are motivated by a clear
inefficiency of the unblocked path search algorithm: it discards all the la-
beling information collected from the construction of each augmenting path.
Since, in a large graph, an augmentation typically has a relatively small ef-
fect on the current flow vector, each augmenting path problem is similar to
the next augmenting path problem. One would thus think that the search
for an augmenting path could be organized to preserve information for use
in subsequent augmentations.
A prime example of an algorithm that cleverly preserves such infor-
mation is the historically important algorithm of Dinic [1970], illustrated
in Figure 3.8. Let us assume for simplicity that each lower arc flow bound
is zero. One possible implementation of the algorithm starts with the zero
flow vector and operates in phases. At the start of each phase, we have a
feasible flow vector x and we construct an acyclic network, called the lay-
ered network , which is partitioned in layers (subsets) of nodes as follows:

Construction of the Layered Network

Layer 0 consists of just the sink node t, and layer k consists of all nodes
i such that the shortest unblocked path from i to t has k arcs. Let
k(i) be the layer number of each node i [k(i) = ∞ if i does not belong
to any layer].
If the source node s does not belong to any layer, there must exist a
saturated cut separating s from t, so the current flow is maximal and
the algorithm terminates. Otherwise, we form the layered network as
follows: we delete all nodes i such that k(i) ≥ k(s) and their incident
arcs, and we delete all remaining arcs except the arcs (i, j) such that
k(i) = k(j) + 1 and xij < cij , or k(j) = k(i) + 1 and xij > 0.
Sec. 3.2 The Ford-Fulkerson Algorithm 131

0/2
2 4 2
0/2 0/2
0/1 0/1 3
1 6 1 6
0/1 4
0/1 0/1
3 5 5
0/2

Initial flows/capacities Layered network for 1st phase

0/2
2 4
1/2 1/2
1/1 1/1 1 2 4
1 6
6
1/1
1/1 0/1 5
3 5
0/2

Flows/capacities after 1st phase Layered network for 2nd phase

1/2
2 4
2/2 2/2
1/1 1/1
1 6
1/1
1/1 0/1
3 5
0/2

Flows/capacities after 2nd phase

Figure 3.8: Illustration of Dinic’s algorithm for the problem shown at the top
left (node 1 is the source and node 6 is the sink).
In the first phase, there are three layers, as shown in the top right figure.
There are three augmentations in the layered network (1 → 2 → 6, 1 → 3 → 6,
and 1 → 4 → 6), and the resulting flows are shown in the middle left figure. In
the second phase, there are four layers, as shown in the bottom right figure. There
is only one augmenting path in the layered network (1 → 2 → 4 → 6), and the
resulting flows are shown in the bottom left figure. The algorithm then terminates
because in constructing the layered network, no augmenting paths from 1 to 6
can be found.
Notice a key property of the algorithm: with each new phase, the layer
number of the source node is strictly increased (from 2 to 3 in this example).
This property shows that the number of phases is at most N − 1.
 132 The Max-Flow Problem Chap. 3

Each phase consists of successively performing augmentations using

only arcs of the layered network constructed at the start of the phase, until
no more augmentations can be performed.
It can be seen that with proper implementation, the layered network
can be constructed in O(A) time. Furthermore, the number of augmen-
tations in each phase is at most A, since each augmentation makes at
least one arc unusable for transferring flow from s to t. Given that the
flow changes of each augmentation require O(N ) time, it follows that each
phase requires O(N A) time. Finally, it can be shown that with each phase,
the layer number k(s) of the source node s increases strictly, so that there
can be at most N − 1 phases (we leave this as Exercise 3.13 for the reader).
It thus follows that the running time of the algorithm is O(N 2 A).
We note that the Dinic algorithm motivated a number of other max-
flow algorithms with improved complexity, including an algorithm of Karza-
nov [1974], which has a O(N 3 ) running time (see the sources cited at the
end of the chapter). The Karzanov algorithm in turn embodied some of
the ideas that were instrumental for the development of the preflow-push
algorithms for max-flow, which will be discussed in Section 7.3.

3.3 PRICE-BASED AUGMENTING PATH ALGORITHMS

In this section, we develop another type of Ford-Fulkerson algorithm, which

reuses information from one augmentation to the next, but does not con-
struct shortest augmenting paths. With proper implementation, this al-
gorithm can be shown to have an O(N 2 A) running time. However, there
is evidence that in practice it outperforms the Dinic and the Karzanov
algorithms, as well as the preflow-push algorithms of Section 7.3.
We mentioned earlier that constructing shortest augmenting paths
provides some guarantee of computational efficiency in the Ford-Fulkerson
algorithm. We can in fact view formally the problem of constructing such
an augmenting path as a shortest path problem in a certain graph, which
we will call the reduced graph. In particular, given a capacity-feasible flow
vector x, this graph has a node set that is the same as the one of the original
graph, and an arc set that is constructed from the one of the original graph
by reversing the direction of some of the arcs and by duplicating some arcs
and then reversing their direction. In particular, it contains:
(a) An arc (i, j) for each arc (i, j) of the original problem’s graph with
xij < cij .
(b) An arc (j, i) for each arc (i, j) of the original problem’s graph with
bij < xij .
Thus each incident arc of a node i (either outgoing or incoming) in the
original graph along which flow can be pushed from i towards the opposite
Sec. 3.3 Price-Based Augmenting Path Algorithms 133

node, corresponds to an outgoing arc from i in the reduced graph. Fur-

thermore, a path in the original graph is unblocked if it corresponds to
a forward path of the reduced graph. Figure 3.9 illustrates the reduced
graph.

ORIGINAL GRAPH
Max-Flow/Capacity
shown next to each arc.
All lower flow bounds are 0.
1/2
2 4
1/1 1/1 Figure 3.9: Illustration of the reduced
graph corresponding to a given flow vec-
0/1 0/1
tor. Node 1 is the source, and node 6 is
1 6 the sink.
0/2
0/1 Figure (a) shows the original graph,
0/1 0/1 and the flow and upper flow bound next
3 5 to each arc (all lower flow bounds are
0/2 0). Figure (b) shows the reduced graph.
The arc (4,2) is added because the flow
(a) of arc (2,4) is strictly between the arc
flow bounds. The arcs (1,2) and (4,6)
REDUCED GRAPH are reversed because their flows are at
the corresponding upper bounds.
Note that every forward path in
2 4
the reduced graph, such as (1, 4, 2, 6),
corresponds to an unblocked path in the
original graph.
1 6

3 5

(b)

It can now be seen that, given a capacity-feasible flow vector, the

problem of finding an augmenting path from s to t with a minimum number
of arcs is equivalent to the problem of finding a shortest path from s to t
in the corresponding reduced graph, with each arc having length 1. This
suggests the simple idea of embedding one of the shortest path algorithms of
Chapter 2 within the Ford-Fulkerson method. The shortest path algorithm
will be used to construct the sequence of augmenting paths from s to t.
Ideally, the algorithm should reuse some information from one shortest
path construction to the next; we mentioned earlier that this is a key to
computational efficiency.
Reusing information for a shortest path method amounts to provid-
134 The Max-Flow Problem Chap. 3

ing some form of advanced initialization, such as label information in the

context of label correcting methods or price information in the context
of auction algorithms. In particular, following a shortest path augmenta-
tion, and the attendant change of the reduced graph, one would like to
be able to reuse at least some of the final data of the preceding shortest
path construction, to provide an advanced start for the next shortest path
construction. Unfortunately, label correcting methods do not seem well
suited for this purpose, because it turns out that following a change of the
reduced graph due to an augmentation, many of the corresponding node
labels can become unusable.
On the other hand, the auction algorithm of Section 2.6 is much
better suited. The reason is that the node prices in the auction algorithm
are required to satisfy the CS condition

pi ≤ p j + 1 (3.6)

for all arcs (i, j) of the reduced graph. Furthermore, upon discovery of a
shortest augmenting path, there holds

pi = p j + 1

for all arcs (i, j) of the augmenting path. It can be seen that this equality
guarantees that following a flow augmentation, the CS condition (3.6) will
be satisfied for all newly created arcs of the reduced graph. As a result,
following an augmentation along a shortest path found by the auction al-
gorithm, the node prices can be reused without modification to start the
auction algorithm for finding the next shortest augmenting path.
The preceding observations can be used to formally define a max-
flow algorithm, where each augmenting path is found as a shortest path
from s to t in the reduced graph using the auction algorithm as a shortest
path subroutine. The initial node prices can be all equal to 0, and the
prevailing prices upon discovery of a shortest augmenting path are used as
the starting prices for searching for the next augmenting path. The auction
algorithm maintains a path starting at s, which is contracted or extended
at each iteration. The price of the terminal node of the path increases by at
least 1 whenever there is a contraction. An augmentation occurs whenever
the terminal node of the path is the sink node t. The overall algorithm is
terminated when the price of the terminal node exceeds N − 1, indicating
that there is no path starting at s and ending at t.
It is possible to show that, with proper implementation, the max-
flow algorithm just described has an O(N 2 A) running time. Unfortunately,
however, the practical performance of the algorithm is not very satisfactory,
because the computation required by the auction/shortest path algorithm
is usually much larger than what is needed to find an augmenting path. The
reason is that one needs just a path from s to t in the reduced graph and
Sec. 3.3 Price-Based Augmenting Path Algorithms 135

insisting on obtaining a shortest path may involve a substantial additional

computational cost. In what follows, we will give a price-based method that
constructs a (not necessarily shortest) path from s to t. This method is
similar to the auction/shortest path algorithm, but when embedded within
a sequential augmenting path construction scheme, it results in a max-flow
algorithm that is much faster in practice.

3.3.1 A Price-Based Path Construction Algorithm

We will describe a special method for finding a simple forward path in a

directed graph (N , A) that starts at a given node s and ends at a given
node t. This method will be subsequently embedded within a max-flow
context to construct augmenting paths. The algorithm maintains (except
upon termination) a simple forward path P = (s, n1 , . . . , nk ) and a set of
integer node prices pi , i ∈ N , satisfying

pi ≤ pj + 1, ∀ (i, j) ∈ A, (3.7)

ps < N, pt = 0, (3.8)

p i ≥ pj , ∀ (i, j) ∈ P. (3.9)

[Note the difference with the auction/shortest path algorithm of Section

2.6, where we require that pi = pj + 1 for all arcs (i, j) of the path P ,
rather than pi ≥ pj .]
At the start of the algorithm, we require that P = (s), and that p
is such that Eqs. (3.7) and (3.8) hold. The path P is modified repeatedly
using the following two operations:
(a) A contraction of P , which deletes the last arc of P , that is, replaces the
path P = (s, n1 , . . . , nk ) by the path P = (s, n1 , . . . , nk−1 ). [In the
degenerate case where P = (s), a contraction leaves P unchanged.]
(b) An extension of P , which adds to P an arc outgoing from its end
node, that is, replaces the path P = (s, n1 , . . . , nk ) by a path P =
(s, n1 , . . . , nk , nk+1 ), where (nk , nk+1 ) is an arc.
The prices pi may also be increased in the course of the algorithm so that,
together with P , they satisfy the conditions (3.7)-(3.9). A contraction
always involves a price increase of the end node nk . An extension may or
may not involve such a price increase. An extension of P is always done to
a neighbor node of nk that has minimal price. The algorithm terminates if
either node t becomes the end node of P (then P is the desired path), or
else ps ≥ N [in view of pt = 0 and pi ≤ pj + 1 for all arcs (i, j), as per Eqs.
(3.7) and (3.8), this means that there is no forward path from s to t].
136 The Max-Flow Problem Chap. 3

Path Construction Algorithm

Set P = (s), and select p such that Eqs. (3.7) and (3.8) hold.
Step 1 (Check for contraction or extension): Let nk be the
end node of the current path P and if nk = s, let pred(nk ) be the
predecessor node of nk on P . If the set of downstream neighbors of
nk ,
N (nk ) = {j | (nk , j) ∈ A},
is empty, set pnk = N and go to Step 3. Otherwise, find a node in
N (nk ) with minimal price and denote it succ(nk ),

succ(nk ) = arg min pj . (3.10)

j∈N (nk )

Set
pnk = psucc(nk ) + 1. (3.11)
If nk = s, or if

nk = s and ppred(nk ) > psucc(nk ) ,

go to Step 2; otherwise go to Step 3.

Step 2 (Extend
 path): Extend P by node succ(nk ) and the corre-
sponding arc nk , succ(nk ) . If succ(nk ) = t, terminate the algorithm;
otherwise go to Step 1.
Step 3 (Contract path): If P = (s) and ps ≥ N , terminate the
algorithm; otherwise, contract P and go to Step 1.

Figure 3.10 illustrates the preceding path construction algorithm. In

the special case where all initial prices are zero and there is a path from each
node to t, by tracing the steps, it can be seen that the algorithm will work
like depth-first search, raising to 1 the prices of the nodes of some path from
s to t in a sequence of extensions with no intervening contractions. More
generally, the algorithm terminates without performing any contractions if
the initial prices satisfy pi ≥ pj for all arcs (i, j) and there is a path from
each node to t.
Note that the algorithm does not necessarily generate a shortest path.
Instead, it can be shown that it solves a special type of assignment problem
by means of the auction algorithm of Section 1.3.3 (which will be further
developed in Chapter 7); see Exercise 3.17.
We make the following observations:
(1) The prices remain integer throughout the algorithm [cf. Eq. (3.11)].
Sec. 3.3 Price-Based Augmenting Path Algorithms 137

p2 = -1 p2 = 1
2 2
p1 = 0 p4 = 0 p1 = 1 p4 = 0

1 4 1 4
Origin Destination

3 3
p3 = 0 p3 = 1

Path construction problem with Trajectory of end node of the

initial prices as shown path P and final prices
generated by the algorithm

Iteration # Path P prior Type of action Price vector p after

to iteration during iteration the iteration

1 (1) extension to 2 (0, −1, 0, 0)

2 (1, 2) contraction at 2 (0, 1, 0, 0)
3 (1) extension to 3 (1, 1, 0, 0)
4 (1, 3) extension to 4 (1, 1, 1, 0)
5 (1, 3, 4) stop

Figure 3.10: An example illustrating the path construction algorithm from s = 1

to t = 4, where the initial price vector is p = (0, −1, 0, 0).

(2) The conditions (3.7)-(3.9) are satisfied each time Step 1 is entered.
The proof is by induction. These conditions hold initially by assump-
tion. Condition (3.8) is maintained by the algorithm, since termi-
nation occurs as soon as ps ≥ N or t becomes the end node of P .
To verify conditions (3.7) and (3.9), we note that only the price of
nk can change in Step 1, and by Eqs. (3.10) and (3.11), this price
change maintains condition (3.7) for all arcs, and condition
 (3.9) for
all arcs of P , except possibly for the arc pred(nk ), nk in the case of
an extension with the condition
ppred(nk ) > psucc(nk )
holding. In the latter case, we must have
ppred(nk ) ≥ psucc(nk ) + 1
because the prices are integer, so by Eq. (3.11), we have
ppred(nk ) ≥ pnk
138 The Max-Flow Problem Chap. 3

at the next entry to Step 1. This completes the induction.

(3) A contraction is always accompanied by a price increase. Indeed
by Eq. (3.9), which was just established, upon entering Step 1 with
nk = s, we have
pnk ≤ ppred(nk ) ,
and to perform a contraction, we must have
ppred(nk ) ≤ psucc(nk ) .
Hence
pnk ≤ psucc(nk ) ,
implying by Eq. (3.11) that p(nk ) must be increased to psucc(nk ) + 1.
It can be seen, however, by example (see Fig. 3.10), that an extension
may or may not be accompanied by a price increase.
(4) Upon return to Step 1 following an extension, the end node nk satisfies
[cf. Eq. (3.11)]
ppred(nk ) = pnk + 1.
This, together with the condition pi ≥ pj for all (i, j) ∈ P [cf. Eq.
(3.9)], implies that the path P will not be extended to a node that
already belongs to P , thereby closing a cycle. Thus P remains a
simple path throughout the algorithm.
The following proposition establishes the termination properties of
the algorithm.

Proposition 3.3: If there exists a forward path from s to t, the

path construction algorithm terminates via Step 2 with such a path.
Otherwise, the algorithm terminates via Step 3 when ps ≥ N .

Proof: We first note that the prices of the nodes of P are upper bounded
by N in view of Eqs. (3.8) and (3.9). Next we observe that there is a
price change of at least one unit with each contraction, and since the prices
of the nodes of P are upper bounded by N , there can be only a finite
number of contractions. Since P never contains a cycle, there can be at
most N − 1 successive extensions without a contraction, so the algorithm
must terminate. Throughout the algorithm, we have pt = 0 and pi ≤ pj + 1
for all arcs (i, j). Hence, if a forward path from s to t exists, we must have
ps < N throughout the algorithm, including at termination, and since
termination via Step 3 requires that ps ≥ N , it follows that the algorithm
must terminate via Step 2 with a path from s to t. If a forward path from
s to t does not exist, termination can only occur via Step 3, in which case
we must have ps ≥ N . Q.E.D.
 Sec. 3.4 Notes, Sources, and Exercises 139

3.3.2 A Price-Based Max-Flow Algorithm

Let us now return to the max-flow problem. We can construct an aug-

menting path algorithm of the Ford-Fulkerson type that uses the path con-
struction algorithm just presented. The algorithm consists of a sequence
of augmentations, each performed using the path construction algorithm
to obtain a path of the reduced graph that starts at the source node s and
ends at the sink node t. As starting price vector we can use the zero vector.
An important point here is that, following an augmentation, the price
vector of the path construction algorithm can remain unchanged . The rea-
son is that the node prices in the path construction algorithm are required
to satisfy the condition
pi ≤ p j + 1 (3.12)
for all arcs (i, j) of the reduced graph. Furthermore, upon discovery of an
augmenting path P , there holds

pi ≥ pj

for all arcs (i, j) of P . It follows that as the reduced graph changes due to
the corresponding augmentation, for every newly created arc (j, i) of the
reduced graph, the arc (i, j) must belong to P , so that pi ≥ pj . Hence the
newly created arc (j, i) of the reduced graph will also satisfy the required
condition pj ≤ pi + 1 [cf. Eq. (3.12)].
For a practically efficient implementation of the max-flow algorithm
just described, a number of fairly complex modifications may be needed. A
description of these and a favorable computational comparison with other
competing methods can be found in Bertsekas [1995c], where an O(N 2 A)
complexity bound is also shown for a suitable variant of the method.

3.4 NOTES, SOURCES, AND EXERCISES

The max-flow/min-cut theorem was independently given in Dantzig and

Fulkerson [1956], Elias, Feinstein, and Shannon [1956], and Ford and Fulk-
erson [1956b]. The material of Section 3.1.4 on decomposition of infeasible
problems is apparently new.
The proof that the Ford-Fulkerson algorithm with breadth-first search
has polynomial complexity O(N A2 ) (Exercise 3.12) is due to Edmonds and
Karp [1972]. Using the idea of a layered network, this bound was improved
to O(N 2 A) by Dinic [1970], whose work motivated a lot of research on max-
flow algorithms with improved complexity. In particular, Dinic’s complex-
ity bound was improved to O(N 3 ) by Karzanov [1974] and by Malhotra,
Kumar, and Maheshwari [1978], to O(N 2 1/2
 A 2) by  Cherkasky [1977], to
5/3 2/3
O(N A ) by Galil [1980], and to O N A log N by Galil and Naamad
140 The Max-Flow Problem Chap. 3

[1980]. Dinic’s algorithm when applied to the maximal matching problem

(Exercise 3.9) can be shown to have running time O(N 1/2 A) (see Hopcroft
and Karp [1973]). The survey paper by Ahuja, Magnanti, and Orlin [1989]
provides a complexity-oriented account of max-flow algorithms.
The max-flow algorithm of Section 3.3 is due to Bertsekas [1995c].
This reference contains several variants of the basic method, a discussion of
implementation issues, and extensive computational results that indicate
a superior practical performance over competing methods, including the
preflow-push algorithms of Chapter 7.
There are two important results in network optimization that deal
with the existence of feasible solutions for minimum cost flow problems.
The first is the feasible distribution theorem, due to Gale [1957] and Hoff-
man [1960], which is a consequence of the max-flow/min-cut theorem (Ex-
ercise 3.3). The second is the feasible differential theorem, due to Minty
[1960], which deals with the existence of a set of prices satisfying certain
constraints. This theorem is a consequence of the duality theory to be
fully developed in Chapter 5, and will be given in Exercise 5.11 (see also
Exercise 5.12).

EXERCISES

3.1

Consider the max-flow problem of Fig. 3.11, where s = 1 and t = 5.

(a) Enumerate all cuts of the form [S, N − S] such that 1 ∈ S and 5 ∈
/ S.
Calculate the capacity of each cut.
(b) Find the maximal and minimal saturated cuts.
(c) Apply the Ford-Fulkerson method to find the maximum flow and verify the
max-flow/min-cut equality.

[0,5] 2 [0,1]
[0,1] Figure 3.11: Max-flow problem
1 [0,5] 4 5
[0,4] for Exercise 3.1. The arc capac-
[0,1]
ities are shown next to the arcs.
[0,2] [0,3]
3
Sec. 3.4 Notes, Sources, and Exercises 141

3.2 (Breadth-First Search)

Let i and j be two nodes of a directed graph (N , A).

(a) Consider the following algorithm, known as breadth-first search, for finding
a path from i to j. Let T0 = {i}. For k = 0, 1, . . ., let

/ ∪kp=0 Tp | for some node m ∈ Tk , (m, n) or (n, m) is an arc},

Tk+1 = {n ∈

and mark each node n ∈ Tk+1 with the label “(m, n)” or “(n, m),” where
m is a node of Tk such that (m, n) or (n, m) is an arc, respectively. The
algorithm terminates if either (1) Tk+1 is empty or (2) j ∈ Tk+1 . Show
that case (1) occurs if and only if there is no path from i to j. If case (2)
occurs, how would you use the labels to construct a path from i to j?
(b) Show that a path found by breadth-first search has a minimum number of
arcs over all paths from i to j.
(c) Modify the algorithm of part (a) so that it finds a forward path from i to
j.

3.3 (Feasible Distribution Theorem)

Show that the minimum cost flow problem introduced in Section 1.2.1, has a
feasible solution if and only if s = 0 and for every cut Q = [S, N − S] we
i∈N i
have 
Capacity of Q ≥ si .
i∈S

Show also that feasibility of the problem can be determined by solving a max-
flow problem with zero lower flow bounds. Hint: Assume first that all lower flow
bounds bij are zero. Use the conversion to a max-flow problem of Fig. 3.1, and
apply the max-flow/min-cut theorem. In the general case, transform the problem
to one with zero lower flow bounds.

3.4 (Finding a Feasible Flow Vector)

Describe an algorithm of the Ford-Fulkerson type for checking the feasibility and
finding a feasible solution of a minimum cost flow problem (cf., Section 1.2.1). If
the supplies si and the arc flow bounds bij and cij are integer, your algorithm
should be guaranteed to find an integer feasible solution (assuming at least one
feasible solution exists). Hint: Use the conversion to a max-flow problem of Fig.
3.1.

3.5 (Integer Approximations of Feasible Solutions)

Given a graph (N , A) and a flow vector x with integer divergence, show that
there exists an integer flow vector x having the same divergence vector as x and
satisfying
|xij − xij | < 1, ∀ (i, j) ∈ A.
142 The Max-Flow Problem Chap. 3

Hint: For each arc (i, j), define the integer flow bounds

bij =
xij , cij = xij .

Use the result of Exercise 3.3.

3.6

Consider a graph with arc flow range [0, cij ] for each arc (i, j), and let x be a
capacity-feasible flow vector.
(a) Consider any subset S of nodes all of which have nonpositive divergence
and at least one of which has negative divergence. Show that there must
exist at least one arc (i, j) with i ∈
/ S and j ∈ S such that xij > 0.
(b) Show that for each node with negative divergence there is an augmenting
path that starts at that node and ends at a node with positive divergence.
Hint: Construct such a path using an algorithm that is based on part (a).

3.7 (Ford-Fulkerson Method Counterexample)

This counterexample (from Chvatal [1983]) illustrates how the version of the Ford-
Fulkerson method where augmenting paths need not have as few arcs as possible
may not terminate for a problem with irrational arc flow bounds. Consider the
max-flow problem shown in Fig. 3.12.
(a) Verify that an infinite sequence of augmenting paths is characterized by
the table of Fig. 3.12; each augmentation increases the divergence out of
the source s but the sequence of divergences converges to a value, which
can be arbitrarily smaller than the maximum flow.
(b) Solve the problem with the Ford-Fulkerson method (where the augmenting
paths involve a minimum number of arcs, as given in Section 3.2).

3.8 (Graph Connectivity – Menger’s Theorem)

Let s and t be two nodes in a directed graph. Use the max-flow/min-cut theorem
to show that:
(a) The maximum number of forward paths from s to t that do not share any
arcs is equal to the minimum number of arcs that when removed from the
graph, eliminate all forward paths from s to t.
(b) The maximum number of forward paths from s to t that do not share any
nodes (other than s and t) is equal to the minimum number of nodes that
when removed from the graph, eliminate all forward paths from s to t.
Sec. 3.4 Notes, Sources, and Exercises 143

1 3 6

s 4

2 5 t

After Iter. # Augm. Path x12 x36 x46 x65

6k + 1 (s, 1, 2, 3, 6, t) σ 1 − σ 3k+2 σ − σ 3k+1 0

6k + 2 (s, 2, 1, 3, 6, 5, t) σ− σ 3k+2 1 σ− σ 3k+1 σ 3k+2
6k + 3 (s, 1, 2, 4, 6, t) σ 1 σ − σ 3k+3 σ 3k+2
6k + 4 (s, 2, 1, 4, 6, 3, t) σ − σ 3k+3 1 − σ 3k+3 σ σ 3k+2
6k + 5 (s, 1, 2, 5, 6, t) σ 1− σ 3k+3 σ σ 3k+4
6k + 6 (s, 2, 1, 5, 6, 4, t) σ − σ 3k+4 1 − σ 3k+3 σ − σ 3k+4 0
6(k + 1) + 1 (s, 1, 2, 3, 6, t) σ 1− σ 3(k+1)+2 σ− σ 3(k+1)+1 0

Figure 3.12: Max-flow problem illustrating that if the augmenting paths in the
Ford-Fulkerson method do not have a minimum number of arcs, then the method
may not terminate. All lower arc flow bounds are zero. The upper flow bounds
are larger than one, with the exception of the thick-line arcs; these are arc (3, 6)
which has upper flow bound equal  to one, and arcs (1, 2) and (4, 6) which have
√ 
upper flow bound equal to σ = − 1 + 5 /2. (Note a crucial property of σ; it
satisfies σ k+2 = σ k − σ k+1 for all integer k ≥ 0.) The table gives a sequence of
augmentations.

3.9 (Max Matching/Min Cover Theorem (König-Egervary))

Consider a bipartite graph consisting of two sets of nodes S and T such that
every arc has its start node in S and its end node in T . A matching is a subset of
arcs such that all the start nodes of the arcs are distinct and all the end nodes of
the arcs are distinct. A maximal matching is a matching with a maximal number
of arcs.
(a) Show that the problem of finding a maximal matching can be formulated
as a max-flow problem.
(b) Define a cover C to be a subset of S ∪ T such that for each arc (i, j), either
i ∈ C or j ∈ C (or both). A minimal cover is a cover with a minimal number
of nodes. Show that the number of arcs in a maximal matching and the
number of nodes in a minimal cover are equal. (Variants of this theorem
were independently published by König [1931] and Egervary [1931].) Hint:
Use the max-flow/min-cut theorem.
144 The Max-Flow Problem Chap. 3

3.10 (Theorem of Distinct Representatives, Hall [1956])

Given finite sets S1 , S2 , . . . , Sk , we say that the collection {s1 , s2 , . . . , sk } is a

system of distinct representatives if si ∈ Si for all i and si = sj for i = j. (For
example, if S1 = {a, b, c}, S2 = {a, b}, S3 = {a}, then s1 = c, s2 = b, s3 = a is a
system of distinct representatives.) Show that there exists no system of distinct
representatives if and only if there exists an index set I ⊂ {1, 2, . . . , k} such
that the number of elements in ∪i∈I Si is less than the number of elements in I.
Hint: Consider a bipartite graph with each of the right side nodes representing
an element of ∪i∈I Si , with each of the left side nodes representing one of the sets
S1 , S2 , . . . Sk , and with an arc from a left node S to a right node s if s ∈ S. Use
the maximal matching/minimal cover theorem of Exercise 3.9. For additional
material on this problem, see Hoffman and Kuhn [1956], and Mendelssohn and
Dulmage [1958].

3.11

Prove the following generalizations of Prop. 3.1:

(a) Let x be a capacity-feasible flow vector, and let N + and N − be two disjoint
subsets of nodes. Then exactly one of the following two alternatives holds:
(1) There exists a simple path that starts at some node of N + , ends at
some node of N − , and is unblocked with respect to x.
(2) There exists a saturated cut Q = [S, N − S] such that N + ⊂ S and
N − ⊂ N − S.
(b) Show part (a) with “simple path” in alternative (1) replaced by “path”.
Hint: Use the path decomposition theorem of Exercise 1.4.

3.12 (Termination of the Ford-Fulkerson Algorithm)

Consider the Ford-Fulkerson algorithm as described in Section 3.2 (augmenting

paths have as few arcs as possible). This exercise shows that the algorithm
terminates and solves the max-flow problem in polynomial time, even when the
problem data are irrational.
Let x0 be the initial feasible flow vector; let xk , k = 1, 2, . . ., be the flow
vector after the kth augmentation; and let Pk be the corresponding augmenting
path. An arc (i, j) is said to be a k+ -bottleneck if (i, j) is a forward arc of Pk and
xkij = cij , and it is said to be a k− -bottleneck if (i, j) is a backward arc of Pk and
xkij = bij .
+
(a) Show that if k < k and an arc (i, j) is a k+ -bottleneck and a k -bottleneck,
then for some m with k < m < k, the arc (i, j) is a backward arc of Pm .
−
Similarly, if an arc (i, j) is a k− -bottleneck and a k -bottleneck, then for
some m with k < m < k, the arc (i, j) is a forward arc of Pm .
(b) Show that Pk is a path with a minimal number of arcs over all augmenting
paths with respect to xk−1 . (This property depends on the implementation
of the unblocked path search as a breadth-first search.)
Sec. 3.4 Notes, Sources, and Exercises 145

(c) For any path P that is unblocked with respect to xk , let nk (P ) be the
number of arcs of P , let a+k (i) be the minimum of nk (P ) over all unblocked
P from s to i, and let a− k (i) be the minimum of nk (P ) over all unblocked
P from i to t. Show that for all i and k we have

a+ +
k (i) ≤ ak+1 (i), a− −
k (i) ≤ ak+1 (i).

+
(d) Show that if k < k and arc (i, j) is both a k+ -bottleneck and a k -
−
bottleneck, or is both a k− -bottleneck and a k -bottleneck, then a+
k (t) <
+
a (t).
k

(e) Show that the algorithm terminates after O(N A) augmentations, for an
O(N A2 ) running time.

3.13 (Layered Network Algorithm)

Consider the algorithm described near the end of Section 3.2, which uses phases
and augmentations through a layered network.
(a) Provide an algorithm for constructing the layered network of each phase in
O(A) time.
(b) Show that the number of augmentations in each phase is at most A, and
provide an implementation whereby these augmentations require O(N A)
total time.
(c) Show that with each phase, the layer number k(s) of the source node s
increases strictly, so that there can be at most N − 1 phases.
(d) Show that with the implementations of (a) and (b), the running time of
the algorithm is O(N 2 A).

3.14 (O(N 2/3 A) Complexity for Unit Capacity Graphs)

Consider the max-flow problem in the special case where the arc flow range is
[0,1] for all arcs.
(a) Show that each path from the source to√the sink that is unblocked with
respect to the zero flow has at most 2N/ M arcs, where M is the value of
the maximum flow. Hint: Let Nk be the number of nodes i such that the
shortest unblocked path from s to i has k arcs. Argue that k(k + 1) ≥ M .
(b) Show that the running time of the layered network algorithm (cf. Fig.
3.8) is reduced to O(N 2/3 A). Hint: Argue that each arc of the layered
network can be part of at most one augmenting path in a given phase, so
the augmentations of each phase require O(A) computation. Use part (a)
to show that the number of phases is O(N 2/3 ).
146 The Max-Flow Problem Chap. 3

3.15

(a) Solve the problem of Exercise 3.1 using the layered network algorithm (cf.
Fig. 3.8).
(b) Construct an example of a max-flow problem where the layered network
algorithm requires N − 1 phases.

3.16

Solve the problem of Exercise 3.1 using the max-flow algorithm of Section 3.3.2.

3.17 (Relation of Path Construction and Assignment)

The purpose of this exercise (from Bertsekas [1995c]) is to show the connection
of the path construction algorithm of Section 3.3.1 with the assignment auction
algorithm of Section 1.3.3.
(a) Show that the path construction problem can be converted into the problem
of finding a solution of a certain assignment problem with all arc values
equal to 0, as shown by example in Fig. 3.13. In particular, a forward path
of a directed graph G that starts at node s and ends at node t corresponds
to a feasible solution of the assignment problem, and conversely.
(b) Show how to relate the node prices in the path construction algorithm with
the object prices of the assignment problem, so that if we apply the auction
algorithm with  = 1, the sequence of generated prices and assignments
corresponds to the sequence of generated prices and paths by the path
construction algorithm.

3.18 (Decomposition of Infeasible Assignment Problems)

Apply the decomposition approach of Section 3.1.4 to an infeasible n × n assign-

ment problem. Show that the set of persons can be partitioned in three disjoint
subsets I1 , I2 , and I3 , and that the set of objects can be partitioned in three
disjoint subsets J1 , J2 , and J3 with the following properties (cf. Fig. 3.14):
(1) I1 , J1 , I2 , and J2 are all nonempty, while I3 and J3 may be empty.
/ I1 and j ∈ J1 , or i ∈ I2 and j ∈
(2) There is no pair (i, j) ∈ A such that i ∈ / J2 .
(3) If I3 and J3 are nonempty, then all pairs (i, j) ∈ A with i ∈ I3 are such
that j ∈ J3 .
Identify the three component problems of the decomposition in terms of the sets
I1 , J1 , I2 , J2 , I3 , and J3 . Show that two of these problems are feasible asymmetric
assignment problems (the numbers of persons and objects are unequal), while the
third is a feasible symmetric assignment problem (the numbers of persons and
objects are equal).
Sec. 3.4 Notes, Sources, and Exercises 147

PERSONS OBJECTS
1
3 t 1

2
1
2' (3,2) 1

s t

1
2 (s,3) 1
3

1
s (s,2) 1

Equivalent Assignment Problem

Figure 3.13: Converting the path construction problem into an equivalent feasi-
bility problem of assigning “persons” to “objects.” Each arc (i, j) of the graph G,
with i
= t, is replaced by an object labeled (i, j). Each node i
= t is replaced by
R(i) persons, where R(i) is the number of arcs of G that are incoming to node i
(for example node 2 is replaced by the two persons 2 and 2 ). Finally, there is one
person corresponding to node s and one object corresponding to node t. For every
arc (i, j) of G, with i
= t, there are R(i) + R(j) incoming arcs from the persons
corresponding to i and j. For every arc (i, t) of G, there are R(i) incoming arcs
from the persons corresponding to i. Each path that starts at s and ends at t can
be associated with a feasible assignment. Conversely, given a feasible assignment,
one can construct an alternating path (a sequence of alternatively assigned and
unassigned pairs) starting at s and ending at t, which defines a path from s to t.

I1

J1

Figure 3.14: Decomposition of

I2 an infeasible assignment problem
(cf. Exercise 3.18).
J2

I3 J3
148 The Max-Flow Problem Chap. 3

3.19 (Perfect Bipartite Matchings)

Consider the problem of matching n persons with n objects on a one-to-one basis

(cf. Exercises 1.21 and 3.9). For each person i there is a given set of objects A(i)
that can be matched with i. A matching is a subset of pairs (i, j) with j ∈ A(i),
such that there is at most one pair for each person and each object. A perfect
matching is one that consists of n pairs, i.e., one where every person is matched
with a distinct object.
(a) Assume that
there exists  a perfect matching. Consider an imperfect match-
ing S = (i, j) | i ∈ I , where I is a set of m < n distinct persons, and


let J = j | there exists i ∈ I with (i, j) ∈ S . Show that given any i ∈ / I,
there exists a sequence {i, j1 , i1 , j2 , i2 , . . . , jk , ik , j} such that j ∈
/ J, the
pairs (i1 , j1 ), . . . , (ik , jk ) belong to S, and j1 ∈ A(i), j2 ∈ A(i1 ), . . . , jk ∈
A(ik−1 ), j ∈ A(ik ). Hint: Use a max-flow formulation, and try to find an
augmenting path in a suitable graph.
(b) Show that there exists a perfect matching if and only if there is no subset
I ⊂ {1, . . . , n} such that the set ∪i∈I A(i) has fewer elements than I.
(c) (König’s Theorem on Perfect Matchings)
Assumethat all the sets A(i),
i = 1, . . . , n, and all the sets B(j) = i | j = A(i) , j = 1, . . . , n, contain
the same number of elements. Show that there exists a perfect matching.

3.20

Consider a feasible max-flow problem. Show that if the upper flow bound of each
arc is increased by α > 0, then the value of the maximum flow is increased by no
more than αA, where A is the number of arcs.

3.21

A town has m dating agencies that match men and women. Agency i has a
list of men and a list of women, and may match a maximum of ci man/woman
pairs from its lists. A person may be in the list of several agencies but may be
matched with at most one other person. Formulate the problem of maximizing
the number of matched pairs as a max-flow problem.

3.22

Consider an n × n chessboard and let A and B be two given squares.

(a) Consider the problem of finding the maximal number of knight paths that
start at A, end at B, and do not overlap, in the sense that they do not
share a square other than A and B. Formulate the problem as a max-flow
problem.
(b) Solve the problem of part (a) using the max-flow algorithm of Section 3.3.2
for the case where n = 8, and the squares A and B are two opposite corners
of the board.
Sec. 3.4 Notes, Sources, and Exercises 149

3.23

Consider the problem of placing n queens on a chessboard of dimensions n × n

so that there is no pair of queens that attack each other.
(a) Formulate the problem of finding a solution as a max-flow problem.
(b) Formulate the problem of counting the number of distinct solutions as a
max-flow problem.

3.24 (Min-Flow Problem)

Consider the “opposite” to the max-flow problem, which is to minimize the di-
vergence out of s over all capacity-feasible flow vectors having zero divergence
for all nodes other than s and t.
(a) Show how to solve this problem by first finding a feasible solution, and by
then using a max-flow algorithm.
(b) Derive an analog to the max-flow/min-cut theorem.
10

Network Problems with

Integer Constraints

Contents

10.1. Formulation of Integer-Constrained Problems

10.2. Branch-and-Bound
10.3. Lagrangian Relaxation
10.3.1. Subgradients of the Dual Function
10.3.2. Subgradient Methods
10.3.3. Cutting Plane Methods
10.3.4. Decomposition and Multicommodity Flows
10.4. Local Search Methods
10.4.1. Genetic Algorithms
10.4.2. Tabu Search
10.4.3. Simulated Annealing
10.5. Rollout Algorithms
10.6. Notes, Sources, and Exercises

467
468 Network Problems with Integer Constraints Chap. 10

In this chapter, we focus again on the general nonlinear network problem

of Chapter 8:
minimize f (x)
subject to x ∈ F,
where x is a flow vector in a given directed graph (N , A), the feasible set
F is
⎧ ⎫
⎨    ⎬

F = x∈X  xij − xji = si , ∀ i ∈ N ,
⎩ ⎭
{j|(i,j)∈A} {j|(j,i)∈A}

and f : F →  is a given real-valued function. Here si are given supply

scalars and X is a given subset of flow vectors. We concentrate on the
case where the feasible set F is discrete because the set X embodies some
integer constraints and possibly some side constraints.
As we noted in Chapter 8, one may solve approximately problems
with integer constraints and side constraints through some heuristic that
neglects in one way or another the integer constraints. In particular, one
may solve the problem as a “continuous” network flow problem and use
some ad hoc method to round the fractional solution to integer. Alterna-
tively, one may discard the complicating side constraints, obtain an integer
solution of the resulting network problem, and use some heuristic to correct
this solution for feasibility of the violated side constraints.
Unfortunately, there are many problems where heuristic methods of
this type are inadequate, and they cannot be relied upon to produce a
satisfactory solution. In such cases, one needs to strengthen the heuristics
with more systematic procedures that can provide some assurance of an
improved solution.
In this chapter we first describe a few examples of integer-constrained
network problems, and we then focus on various systematic solution meth-
ods. In particular, in Section 10.2, we discuss the branch-and-bound method,
which is in principle capable of producing an exactly optimal solution to an
integer-constrained problem. This method relies on upper and lower bound
estimates of the optimal cost of various problems that are derived from the
given problem. Usually, the upper bounds are obtained with various heuris-
tics, while the lower bounds are obtained through integer constraint relax-
ation or through the use of duality. A popular method for obtaining lower
bounds, the Lagrangian relaxation method, is introduced in Section 10.3.
This method requires the optimization of nondifferentiable functions, and
two of the major algorithms that can be used for this purpose, subgradient
and cutting plane methods, are discussed in Section 10.3.
Unfortunately, the branch-and-bound method is too time-consuming
for exact optimal solution, so in many practical problems it can only be
used as an approximation scheme. There are alternative possibilities, which
do not offer the theoretical guarantees of branch-and-bound, but are much
 Sec. 10.1 Formulation of Integer-Constrained Problems 469

faster in practice. Two possibilities of this type, local search methods and
rollout algorithms, are discussed in Sections 10.4 and 10.5, respectively.

10.1 FORMULATION OF INTEGER-CONSTRAINED PROBLEMS

There is a very large variety of integer-constrained network flow problems.

Furthermore, small changes in the problem formulation can often make a
significant difference in the character of the solution. As a result, it is not
easy to provide a taxonomy of the major problems of interest. It is helpful,
however, to study in some detail a few representative examples that can
serve as paradigms when dealing with other problems that have similar
structure. We have already discussed in Section 8.4 an example, the con-
strained shortest path problem. In this section, we provide some additional
illustrative examples of broad classes of integer-constrained problems. In
the exercises, we discuss several variants of these problems.

Example 10.1. Traveling Salesman Problem

An important model for scheduling a sequence of operations is the classical

traveling salesman problem. This is perhaps the most studied of all combi-
natorial optimization problems. In addition to its use as a practical model, it
has served as a testbed for a large variety of formal and heuristic approaches
in discrete optimization.
In a colloquial description of the problem, a salesman wants to find a
minimum mileage/cost tour that visits each of N given cities exactly once
and returns to the city he started form. We associate a node with each city
i = 1, . . . , N , and we introduce an arc (i, j) with traversal cost aij for each
ordered pair of nodes i and j. Note that we assume that the graph is complete;
that is, there exists an arc for each ordered pair of nodes. There is no loss of
generality in doing so because we can assign a very high cost aij to an arc (i, j)
that is precluded from participation in the solution. We allow the possibility
that aij = aji . Problems where aij = aji for all i and j are sometimes called
symmetric or undirected traveling salesman problems, because the direction
of traversal of a given arc does not matter.
A tour (also called a Hamiltonian cycle; see Section 1.1) is defined to be
a simple forward cycle that contains all the nodes of the graph. Equivalently,
a tour is a connected subgraph that consists of N arcs, such that there is
exactly one incoming and one outgoing arc for each node i = 1, . . . , N . If we
define the cost of a subgraph T to be the sum of the traversal costs of its arcs,

aij ,
(i,j)∈T

the traveling salesman problem is to find a tour of minimum cost.

470 Network Problems with Integer Constraints Chap. 10

We formulate this problem as
a network flow problem with  node set

N = {1, . . . , N } and arc set A = (i, j) | i, j = 1, . . . , N, i = j , and with
side constraints and 0-1 integer constraints:

minimize aij xij
(i,j)∈A

subject to xij = 1, i = 1, . . . , N,
j=1,...,N
j=i

xij = 1, j = 1, . . . , N,
i=1,...,N
i=j

xij = 0 or 1, ∀ (i, j) ∈ A,
 
the subgraph with node-arc set N , {(i, j) | xij = 1} is connected. (10.1)
Note that, given the 0-1 constraints on the arc flows and the conservation of
flow equations, the last constraint can be expressed through the set of side
constraints

(xij + xji ) ≥ 2, ∀ nonempty proper subsets S of nodes.
i∈S, j ∈S
/

If these constraints were not present, the problem would be an ordinary as-
signment problem. Unfortunately, however, these constraints are essential,
since without them, there would be feasible solutions involving multiple dis-
connected cycles, as illustrated in Fig. 10.1.

Figure 10.1: Example of an infeasible

1 4 7 solution of a traveling salesman prob-
lem where all the constraints are sat-
isfied except for the connectivity con-
straint (10.1). This solution may have
6 3 been obtained by solving an N × N as-
signment problem and consists of mul-
tiple cycles [(1,2,3), (4,5,6), and (7,8)
in the figure]. The arcs of the cycles
2 5 8 correspond to the assigned pairs (i, j)
in the assignment problem.

A simple approach for solving the traveling salesman problem is the

nearest neighbor heuristic. We start from a path consisting of just a single
node i1 and at each iteration, we enlarge the path with a node that does not
close a cycle and minimizes the cost of the enlargement. In particular, after
k iterations, we have a forward path {i1 , . . . , ik } consisting of distinct nodes,
and at the next iteration, we add an arc (ik , ik+1 ) that minimizes aik i over all
arcs (ik , i) with i = i1 , . . . , ik . After N − 1 iterations, all nodes are included
in the path, which is then converted to a tour by adding the final arc (iN , i1 ).
Sec. 10.1 Formulation of Integer-Constrained Problems 471

Given a tour, one may try to improve its cost by using some method
that changes the tour incrementally. In particular, a popular method for the
symmetric case (aij = aji for all i and j) is the k-OPT heuristic, which creates
a new tour by exchanging k arcs of the current tour with another k arcs that
do not belong to the tour (see Fig. 10.2). The k arcs are chosen to optimize
the cost of the new tour with O(N k ) computation. The method stops when
no improvement of the current tour is possible through a k-interchange.

i j

Figure 10.2: Illustration of the 2-OPT heuristic

for improving a tour of the symmetric traveling
salesman problem. The arcs (i, j) and (i, j) are
interchanged with the arcs (i, j) and (i, j). The
choice of (i, j) and (i, j) is optimized over all pairs
of nonadjacent arcs of the tour.

i j

Another possibility for constructing an initial tour is the following two-

step method:
(1) Discard the side constraints (10.1), and from the resulting assignment
problem, obtain a solution consisting of a collection of subtours such
as the ones shown in Fig. 10.1. More generally, use some method to
obtain a “reasonable” collection of subtours such that each node lies on
exactly one subtour.
(2) Use some heuristic to create a tour by combining subtours. For example,
any two subtours T and T can be merged into a single subtour by
selecting a node i ∈ T and a node i ∈ T , adding the arc (i, i), deleting
the unique outgoing arc (i, j) of i on the subtour T and the unique
incoming arc (j, i) of j on the subtour T , and finally adding the arc
(j, j), as shown in Fig. 10.3. The pair of nodes i and i can be chosen
to minimize the cost of the created subtour. This optimization requires
O(mn) computation, where m and n are the numbers of nodes in T
and T , respectively.
Still another alternative for constructing an initial tour, is to start with
some spanning tree and to gradually convert it into a tour. There are quite a
few heuristics based on this idea; see e.g., the book by Nemhauser and Wolsey
[1988], the survey by Junger, Reinelt, and Rinaldi [1995], and the references
quoted there. Unfortunately, there are no heuristics with practically useful
performance guarantees for the general traveling salesman problem (Sahni
and Gonzalez [1976], and Johnson and Papadimitriou [1985] make this point
precise). The situation is better, however, for some special types of symmetric
472 Network Problems with Integer Constraints Chap. 10

j j Figure 10.3: Merging two subtours

T and T into a single subtour by se-
lecting two nodes i ∈ T and i ∈ T ,
i i
and adding and deleting the appro-
priate arcs of T and T .

Subtour T Subtour T

problems where the arc costs satisfy the relation

aij ≤ aik + akj , for all nodes i, j, k.

known as the triangle inequality (see Exercises 10.7-10.8).

Example 10.2. Fixed Charge Problems

A fixed charge problem is a minimum cost flow problem where there is an

extra cost bij for each arc flow xij that is positive (in addition to the usual
cost aij xij ). Thus bij may be viewed as a “purchase cost” for acquiring the
arc (i, j) and using it to carry flow.
An example of a fixed charge problem is the facility location problem,
where we must select a subset of locations from a given candidate set, and
place in each of these locations a “facility” that will serve the needs of certain
“clients.” There is a 0-1 decision variable associated with selecting any given
location for facility placement, at a given cost. Once these variables are cho-
sen, an assignment (or transportation) problem must be solved to optimally
match clients with facilities. Mathematically, we assume that there are m
clients and n locations. By xij = 1 (or xij = 0) we indicate that client i is
assigned to location j at a cost aij (or is not assigned, respectively). We also
introduce a 0-1 integer variable yj to indicate (with yj = 1) that a facility is
placed at location j at a cost bj . The problem is
 
n

minimize aij xij + b j yj

(i,j)∈A j=1

subject to xij = 1, i = 1, . . . , m,
{j|(i,j)∈A}

xij ≤ yj cj , j = 1, . . . , n,
{i|(i,j)∈A}

xij = 0 or 1, ∀ (i, j) ∈ A,
yj = 0 or 1, j = 1, . . . , n,

where cj is the maximum number of customers that can be served by a facility

at location j.
Sec. 10.1 Formulation of Integer-Constrained Problems 473

We can formulate this problem as a network flow problem with side

constraints and integer constraints. In particular, we can view xij as the arc
flows of the graph of a transportation problem (with inequality constraints).
We can also view yj as the arc flows of an artificial graph that is disconnected
from the transportation graph, but is coupled to it through the side con-
straints x ≤ yj cj (see Fig. 10.4). This formulation does not necessarily
i ij
facilitate the algorithmic solution of the problem, but serves to illustrate the
generality of our framework for network problems with side constraints.

1
1
1 1 1
...

...

...

...
1 xij m yj
i j j j
a ij bj
...

...

...
...

1 n n n
m

CLIENTS LOCATIONS

Figure 10.4: Formulation of the facility location problem as a network flow

problem with side constraints and 0-1 integer constraints. There are two
disconnected subgraphs: the first is a transportation-like graph that involves
the flow variables xij and the second is an artificial graph that involves the
flow variables yj
. The arc flows of the two subgraphs are coupled through the
side constraints x ≤ yj cj .
i ij

Example 10.3. Optimal Tree Problems

There are many network applications where one needs to construct an optimal
tree subject to some constraints. For example, in data networks, a spanning
tree is often used to broadcast information from some central source to all
the nodes. In this context, it makes sense to assign a cost or weight aij to
each arc (communication link) (i, j) and try to find a spanning tree that has
minimum total weight (minimum sum of arc weights). This is the minimum
weight spanning tree problem, which we have briefly discussed in Chapter 2
(see Exercise 2.30).
We can formulate this problem as an integer-constrained problem in
several ways. For example, let xij be a 0-1 integer variable indicating whether
arc (i, j) belongs to the spanning tree. Then the problem can be written as

minimize aij xij
(i,j)∈A

subject to xij = N − 1,
(i,j)∈A
474 Network Problems with Integer Constraints Chap. 10

(xij + xji ) ≥ 1, ∀ nonempty proper subsets S of nodes,
i∈S, j ∈S
/

xij = 0 or 1, ∀ (i, j) ∈ A.
The first two constraints guarantee that the graph defined by the set {(i, j) |
xij = 1} has N − 1 arcs and is connected, so it is a spanning tree.
In Exercise 2.30, we discussed how the minimum weight spanning tree
problem can be solved with a greedy algorithm. An example is the Prim-
Dijkstra algorithm, which builds an optimal spanning tree by generating a
sequence of subtrees. It starts with a subtree consisting of a single node
and it iteratively adds to the current subtree an incident arc that has min-
imum weight over all incident arcs that do not close a cycle. We indicated
in Exercise 2.30 that this algorithm can be implemented so that it has an
O(N 2 ) running time. This is remarkable, because except for the minimum
cost flow problems discussed in Chapters 2-7, very few other types of network
optimization problems can be solved with a polynomial-time algorithm.
There are a number of variations of the minimum weight spanning tree
problem. Here are some examples:
(a) There is a constraint on the number of tree arcs that are incident to
a single given node. This is known as the degree constrained minimum
weight spanning tree problem. It is possible to solve this problem using a
polynomial version of the greedy algorithm (see Exercise 10.10). On the
other hand, if there is a degree constraint on every node, the problem
turns out to be much harder. For example, suppose that the degree of
each node is constrained to be at most 2. Then a spanning tree subject
to this constraint must be a path that goes through each node exactly
once, so the problem is essentially equivalent to a symmetric traveling
salesman problem (see Exercise 10.6).
(b) The capacitated spanning tree problem. Here the arcs of the tree are to
be used for routing specified supplies from given supply nodes to given
demand nodes. The tree specifies the routes that will carry the flow
from the supply points to the demand points, and hence also specifies
the corresponding arc flows. We require that the tree is selected so
that the flow of each arc does not exceed a given capacity constraint.
This is an integer-constrained problem, which is not polynomially solv-
able. However, there are some practical heuristic algorithms, such as
an algorithm due to Esau and Williams [1966] (see Fig. 10.5).
(c) The Steiner tree problem, where the requirement that all nodes must be
included in the tree is relaxed. Instead, we are given a subset S of the
nodes, and we want to find a tree that includes the subset S and has
minimum total weight. [J. Steiner (1796-1863), “the greatest geome-
ter since Apollonius,” posed the problem of finding the shortest tree
spanning a given set of points on the plane.] An important application
of the Steiner tree problem arises in broadcasting information over a
communication network from a special node to a selected subset S of
nodes. This broadcasting is most efficiently done over a Steiner tree,
where the cost of each arc corresponds to the cost of communication
over that arc. The Steiner tree problem also turns out to be a difficult
Sec. 10.1 Formulation of Integer-Constrained Problems 475

0 0
5 5
5 2 3
1 4 1 4
9 9
1 2
1 2
2 3 2 3
1

Spanning Tree Problem Optimal Spanning Tree

Arc Capacities = 8

0 0 0

1 4 1 4 1 4

2 3 2 3 2 3

Starting Tree After 1 Iteration After 2 Iterations

(Final Solution)

Figure 10.5: The Esau-Williams heuristic for solving a capacitated minimum

weight spanning tree problem. Each arc (i, j) has a cost (or weight) aij and a
capacity cij . The problem is symmetric, so that aij = aji and cij = cji . We
assume that the graph is complete [if some arcs (i, j) do not exist, we introduce
them artificially with a very large cost and infinite capacity]. There is a special
concentrator node 0, and for every other node i = 1, . . . , N , there is a supply
si ≥ 0 that must be transferred to node 0 along the arcs of the spanning tree
without violating the arc capacity constraints. The Esau-Williams algorithm
generates a sequence of feasible spanning trees, each having a lower cost than
its predecessor, by using an arc exchange heuristic. In particular, we start
with a spanning tree where the concentrator node 0 is directly connected with
each of the N other nodes, as in the bottom left figure [we assume that the
arcs (i, 0) can carry at least the supply of node i, that is, ci0 ≥ si ]. At each
successive iteration, an arc (i, 0) is deleted from the current spanning tree,
and another arc (i, j) is added, so that:
(1) No cycle is formed.
(2) The capacity constraints of all the arcs of the new spanning tree are
satisfied.
(3) The saving ai0 − aij in cost obtained by exchanging arcs (i, 0) and (i, j)
is positive and is maximized over all nodes i and and j for which (1)
and (2) above are satisfied.
The figure illustrates the algorithm, for the problem shown at the top left,
where the cost of each arc is shown next to each arc, the capacity of each
arc is 8, and the supplies of the nodes i > 0 are shown next to the arrows.
The algorithm terminates after two iterations with the tree shown, which has
a total cost of 13. Termination occurs because when arc (1, 0) or (4, 0) is
removed and an arc that is not incident to node 0 is added, some arc capacity
is violated. The optimal spanning tree has cost equal to 12.
476 Network Problems with Integer Constraints Chap. 10

integer-constrained problem, for which, however, effective heuristics are

available (see Exercise 10.11). Note that there are degree-constrained
and capacitated versions of the problem, as in (a) and (b) above.

Example 10.4. Matching Problems

A matching problem involves dividing a collection of objects into pairs. There

may be some constraints regarding the objects that can be paired, and there
is a benefit or value associated with matching each of the eligible pairs. The
objective is to find a matching of maximal total value. We have already stud-
ied extensively special cases of matching, namely the assignment problems
of Chapter 7, which are also called bipartite matching problems. These are
matching problems where the objects are partitioned in two groups, and pairs
must involve only one element from each group. Matching problems where
there is no such partition are called nonbipartite.
To pose a matching problem as a network flow problem, we introduce
a graph (N , A) that has a node for each object, and an arc (i, j) of value aij
connecting any two objects i and j that can be paired. The orientation of
this arc does not matter [alternatively, we may introduce both arcs (i, j) and
(j, i), and assign to them equal values]. We consider a flow variable xij for
each arc (i, j), where xij is 1 or 0 depending on whether objects i and j are
matched or not, respectively. The objective is to maximize

aij xij
(i,j)∈A

subject to the constraints

 
xij + xji ≤ 1, ∀ i ∈ N, (10.2)
{j|(i,j)∈A} {j|(j,i)∈A}

xij = 0 or 1, ∀ (i, j) ∈ A.

The constraint (10.2) expresses the requirement that an object can be matched
with at most one other object. In a variant of the problem, it is specified that
the matching should be perfect; that is, every object should be matched with
some other object. In this case, the constraint (10.2) should be changed to
 
xij + xji = 1, ∀ i ∈ N. (10.3)
{j|(i,j)∈A} {j|(j,i)∈A}

The special case where aij = 1 for all arcs (i, j) is the maximum cardinal-
ity matching problem, i.e., finding a matching with a maximum number of
matched pairs.
It is possible to view nonbipartite matching as an optimal network flow
problem of the assignment type with integer constraints and with the side con-
straints defined by Eq. (10.2) or Eq. (10.3) (see Exercise 10.15). We would
Sec. 10.1 Formulation of Integer-Constrained Problems 477

thus expect that the problem is a difficult one, and that it is not polynomi-
ally solvable (cf. the discussion of Section 8.4). However, this is not so. It
turns out that nonbipartite matching has an interesting and intricate struc-
ture, which is quite unique among combinatorial and network optimization
problems. In particular, nonbipartite matching problems can be solved with
polynomial-time algorithms. These algorithms share some key structures with
their bipartite counterparts, such as augmenting paths, but they generally be-
come simpler and run faster when specialized to bipartite matching. One such
algorithm, due to Edmonds [1965] can be implemented so that it has O(N 3 )
running time. Furthermore, nonbipartite matching can be formulated as a
linear program without integer constraints, and admits an analysis based on
linear programming duality. We refer to the literature cited at the end of the
chapter for an account.

Example 10.5. Vehicle Routing Problems

In vehicle routing problems, there is a fleet of vehicles that must pick up a

number of “customers” (e.g., persons, packages, objects, etc.) from various
nodes in a transportation network and deliver them at some other nodes using
the network arcs. The objective is to minimize total cost subject to a variety
of constraints. The cost here may include, among other things, transportation
cost, and penalties for tardiness of pickup and delivery. The constraints may
include vehicle capacity, and pickup and delivery time restrictions.
Vehicle routing problems are among the hardest integer programming
problems because they tend to have a large number of integer variables, and
also because they involve both a resource allocation and a scheduling aspect.
In particular, they combine the difficult combinatorial aspects of two problems
that we have already discussed:
(a) The generalized assignment problem discussed in Section 8.5 (determine
which vehicles will service which customers).
(b) The traveling salesman problem discussed in Example 10.1 (determine
the sequence of customer pickups and deliveries by a given vehicle).
In fact, the traveling salesman problem may itself be viewed as a sim-
ple version of the vehicle routing problem, involving a single vehicle
of unlimited capacity, N customers that must be picked up in some
unspecified order, and a travel cost aij from customer i to customer j.
For a common type of vehicle routing problem, suppose that there are
K vehicles (denoted 1, . . . , K) with corresponding capacities c1 , . . . , cK , which
make deliveries to N customers (nodes 1, . . . , N ) starting from a central depot
(node 0). The delivery to customer i is of given size di , and the cost of
traveling from node i to node j is denoted by aij . The problem is to find the
route of each vehicle (a cycle of nodes starting from node 0 and returning to
0), that satisfies the customer delivery constraints, and the vehicle capacity
constraints.
There are several heuristic approaches for solving this problem, some
of which bear similarity to the heuristic approaches for solving the traveling
salesman problem. For example, one may start with some set of routes, which
478 Network Problems with Integer Constraints Chap. 10

may be infeasible because their number may exceed the number of vehicles K.
One may then try to work towards feasibility by combining routes in a way
that satisfies the vehicle capacity constraints, while keeping the cost as small
as possible. Alternatively, one may start with a solution of a K-traveling
salesmen problem (see Exercise 10.9), corresponding to the K vehicles, and
then try to improve on this solution by interchanging customers between
routes, while trying to satisfy the capacity constraints. These heuristics often
work well, but generally they offer no guarantee of good performance, and
may occasionally result in a solution that is far from optimal.
An alternative possibility, which is ultimately also based on heuristics,
is to formulate the problem mathematically in a way that emphasizes its
connections to both the generalized assignment problem and the traveling
salesman problem. In particular, we introduce the integer variables

1 if node i is visited by vehicle k,
yik =
0 otherwise,

and the vectors yk = (y1k , . . . , yN k ). For each k = 1, . . . , K, let fk (yk ) denote

the optimal cost of a traveling salesman problem involving the set of nodes

Nk (yk ) = {i | yik = 1}.

We can pose the problem as


K

minimize fk (yk )
k=1


K 
K if i = 0,
subject to yik =
1 if i = 1, . . . , N,
k=1


N

di yik ≤ ck , k = 1, . . . , K,
i=0

yik = 0 or 1, i = 0, . . . , N, k = 1, . . . , N,

which is a generalized assignment problem (see Section 8.5).

The difficulty with the generalized assignment formulation is that the
functions fk are generally unknown. It is possible, however, to try to approx-
imate heuristically these functions with some linear functions of the form


N

f˜k (yk ) = wik yik ,

i=0

solve the corresponding generalized assignment problems for the vectors yk ,

and then solve the corresponding traveling salesman problems. The weights
wik can be determined in some heuristic way. For example, first specify a
“seed” customer ik to be picked up by vehicle k, and then set

wik = a0i + aiik − a0ik ,

Sec. 10.1 Formulation of Integer-Constrained Problems 479

which is the incremental cost of inserting customer i into the route 0 → ik →

0. The seed customers specify the general direction of the route taken by
vehicle k, and the weight wik represents the approximate cost for picking up
customer i along the way. One may select the seed customers using one of a
number of heuristics, for which we refer to the literature cited at the end of
the chapter.
There are several extensions and more complex variants of the preceding
vehicle routing problems. For example:
(a) Some of the customers may have a “time window,” in the sense that
they may be served only within a given time interval. Furthermore, the
total time duration of a route may be constrained.
(b) There may be multiple depots, and each vehicle may be restricted to
start from a given subset of the depots.
(c) Delivery to some of the customers may not be required. Instead there
may be a penalty for nondelivery or for tardiness of delivery (in the
case where there are time windows).
(d) There may be precedence constraints, requiring that some of the cus-
tomers be served before some others.
With additional side constraints of the type described above, the prob-
lem can become very complex. Nonetheless, with a combination of heuristics
and the more formal approaches to be described in this chapter, some measure
of success has been obtained in solving practical vehicle routing problems.

Example 10.6. Arc Routing Problems

Arc routing problems are similar to vehicle routing problems, except that
the emphasis regarding cost and constraints is placed on arc traversals rather
than node visits. Here each arc (i, j) has a cost aij , and we want to find a
set of arcs that satisfy certain constraints and have minimum sum of costs.
For example, a classical arc routing problem is the Chinese postman problem,
where we want to find a cycle that traverses every arc of a graph, and has
minimum sum of arc costs; here traversals in either direction and multiple
traversals are allowed.† The costs of all arcs must be assumed nonnegative
here in order to guarantee that the problem has an optimal solution (otherwise
cycles of arbitrarily small cost would be possible by crossing back and forth
an arc of negative cost).
An interesting related question is whether there exists an Euler cycle
in the given graph, i.e., a cycle that contains every arc exactly once, with
arc traversals in either direction allowed (such a cycle, if it exists, solves the
Chinese postman problem since the arc costs are assumed nonnegative). This

† An analogy here is made with a postman who must traverse each arc of the
road network of some town (in at least one direction), while walking the minimum
possible distance. The problem was first posed by the Chinese mathematician
Kwan Mei-Ko [1962].
480 Network Problems with Integer Constraints Chap. 10

question was posed by Euler in connection with the famous Königsberg bridge
problem (see Fig. 10.6). The solution is simple: there exists an Euler cycle
if and only if the graph is connected and every node has even degree (in an
Euler cycle, the number of entrances to a node must be equal to the number
of exits, so the number of incident arcs to each node must be even; for a proof
of the converse, see Exercise 1.5). It turns out that even when there are nodes
of odd degree, a solution to the Chinese postman problem can be obtained
by constructing an Euler cycle in an expanded graph that involves some ad-
ditional arcs. These arcs can be obtained by solving a nonbipartite matching
problem involving the nodes of odd degree (see Exercise 10.17). Thus, since
the matching problem can be solved in polynomial time as noted in Example
10.4, the Chinese postman problem can also be solved in polynomial time
(see also Edmonds and Johnson [1973], who explored the relation between
matching and the Chinese postman problem).

C
B River Pregel

C
B

Figure 10.6: The Königsberg bridge problem, generally considered to mark

the origin of graph theory. Euler attributed this problem to the citizens of
Königsberg, an old port town that lies north of Warsaw on the Baltic sea
(it is now called Kaliningrad). The problem, addressed by Euler in 1736, is
whether it is possible to cross each of the seven bridges of the river Pregel
in Königsberg exactly once, and return to the starting point. In the graph
representation of the problem, shown in the figure, each bridge is associated
with an arc, and each node is associated with a land area that is incident
to several bridges. The question amounts to asking whether an Euler cycle
exists. The answer is negative since there are nodes with odd degree.

There is also a “directed” version of the Chinese postman problem,

where we want to find a forward cycle that traverses every arc of a graph
(possibly multiple times), and has minimum sum of arc costs. It can be seen
that this problem has a feasible solution if and only if the graph is strongly
connected, and that it has an optimal solution if in addition all forward
cycles have nonnegative cost. The problem is related to the construction
Sec. 10.1 Formulation of Integer-Constrained Problems 481

of forward Euler cycles, in roughly the same way as the undirected Chinese
postman problem was related above to the construction of an (undirected)
Euler cycle. Exercise 1.8 states the basic result about the existence of a
forward Euler cycle: such a cycle exists if and only if the number of incoming
arcs to each node is equal to the number of its outgoing arcs. A forward Euler
cycle, if it exists, is also a solution to the directed Chinese postman problem.
More generally, it turns out that a solution to the directed Chinese postman
problem (assuming one exists) can be obtained by finding a directed Euler
cycle in an associated graph obtained by solving a certain minimum cost flow
problem (see Exercise 10.17).
By introducing different constraints, one may obtain a large variety of
arc routing problems. For example, a variant of the Chinese postman problem
is to find a cycle of minimum cost that traverses only a given subset of the arcs.
This is known as the rural postman problem. Other variants are characterized
by arc time-windows and arc precedence constraints, similar to vehicle routing
problem variants discussed earlier. In fact, it is always possible to convert
an arc routing problem to a “node routing problem,” where the constraints
are placed on some of the nodes rather than on the arcs. This can be done
by replacing each arc (i, j) with two arcs (i, kij ) and (kij , j) separated by an
artificial middle node kij . Traversal of an arc (i, j) then becomes equivalent
to visiting the artificial node kij . However, this transformation often masks
important characteristics of the problem. For example it would be awkward
to pose the question of existence of an Euler cycle as a node routing problem.

Example 10.7. Multidimensional Assignment Problems

In the assignment problems we have considered so far, we group the nodes of

the graph in pairs. Multidimensional assignment problems involve the group-
ing of the nodes in subsets with more than two elements, such as triplets or
quadruplets of nodes. For an example of a 3-dimensional assignment problem,
suppose that the performance of a job j requires a machine m and a worker
w, and that there is a given value ajmw corresponding to the triplet (j, m, w).
Given a set of jobs J, a set of machines M , and a set of workers W , we want
to find a collection of job/machine/worker triplets that has maximum total
value.
To pose this problem mathematically, we introduce 0-1 integer variables

1 if job j is performed at machine m by worker w,
xjmw =
0 otherwise,
and we maximize  
ajmw xjmw
j∈J m∈M w∈W

subject to standard assignment constraints. In particular, if the numbers of

jobs, machines, and workers are all equal, and all jobs must be assigned, we
have the constraints
 
xjmw = 1, ∀ j ∈ J,
m∈M w∈W
482 Network Problems with Integer Constraints Chap. 10

xjmw = 1, ∀ m ∈ M,
j∈J w∈W

xjmw = 1, ∀ w ∈ W.
j∈J m∈M

In alternative formulations, some of these constraints may involve inequalities.

An important and particularly favorable special case of the problem
arises when the values ajmw have the separable form

ajmw = βjm + γmw ,

where βjm and γmw are given scalars. In this case, there is no coupling
between jobs and workers, and the problem can be solved by solving two
decoupled (2-dimensional) assignment problems: one involving the pairing of
jobs and machines, with the βjm as values, and the other involving the pairing
of machines and workers, with the γmw as values. In general, however, the 3-
dimensional assignment problem is a difficult integer programming problem,
for which there is no known polynomial algorithm.
A simple heuristic approach is based on relaxing each of the constraints
in turn. In particular, suppose that the constraint on the workers is neglected
first. It can then be seen that the problem takes the 2-dimensional assignment
form

maximize bjm yjm
j∈J m∈M

subject to yjm = 1, ∀ j ∈ J,
m∈M

yjm = 1, ∀ m ∈ M,
j∈J

yjm = 0 or 1, ∀ j ∈ J, m ∈ M,

where
bjm = max ajmw , (10.4)
w∈W

and yjm = 1 indicates that job j must be performed at machine m. For each
j ∈ J, let jm be the job assigned to machine m, according to the solution of
this problem. We can now optimally assign machines m to workers w, using
as assignment values
cmw = ajm mw ,

and obtain a 3-dimensional assignment {(jm , m, wm ) | m ∈ M }. It can be

seen that this approach amounts to enforced separation, whereby we replace
the values ajmw with the separable approximations bjm + cmw . In fact, it
can be shown that if the problem is -separable, in the sense that for some
(possibly unknown) β jm and γ mw , and some  ≥ 0, we have

|β jm + γ mw − ajmw | ≤ , ∀ j ∈ J, m ∈ M, w ∈ W,
 Sec. 10.2 Branch-and-Bound 483

then the assignment {(jm , m, wm ) | m ∈ M } obtained using the preceding

enforced separation approach achieves the optimal value of the problem within
4n, where n is the cardinality of the sets J, M , and W (see Exercise 10.31).
The enforced separation approach is simple and can be generalized to
multidimensional assignment problems of dimension more than 3. However,
it often results in significant loss of optimality. A potential improvement is
to introduce some corrections to the values bjm that reflect some dependence
on the values of workers. For example, we can use instead of the values bjm
of Eq. (10.4), the modified values

b̂jm = max {ajmw − µw },

w∈W

where µw is a nonnegative scalar that can be viewed as a wage to be paid

to worker w. This allows the possibility of adjusting the scalars µw to some
“optimal” values. Methods for doing this will be discussed in Section 10.3 in
the context of the Lagrangian relaxation method, where
we will view µw as a
Lagrange multiplier corresponding to the constraint j∈J m∈M xjmw = 1.
There are several extensions of the multidimensional assignment prob-
lem. For example, we may have transportation constraints, where multiple
jobs can be performed on the same machine, and/or multiple machines can
be operated by a single worker. In this case, our preceding discussion of the
enforced separation heuristic applies similarly. We may also have generalized
assignment constraints such as

gjmw xjmw ≤ 1, ∀ m ∈ M,
j∈J w∈W

where gjmw represents the portion of machine m needed to perform job j by

worker w. In this case, the enforced separation heuristic results in difficult
integer-constrained generalized assignment problems, which we may have to
solve heuristically. Alternatively, we may use the more formal methodology
of the next two sections.

10.2 BRANCH-AND-BOUND

The branch-and-bound method implicitly enumerates all the feasible so-

lutions, using calculations where the integer constraints of the problem
are relaxed. The method can be very time-consuming, but is in principle
capable of yielding an exactly optimal solution.
To describe the branch-and-bound method, consider the general dis-
crete optimization problem

minimize f (x)
subject to x ∈ F,
484 Network Problems with Integer Constraints Chap. 10

where the feasible set F is a finite set. The branch-and-bound algorithm

uses an acyclic graph known as the branch-and-bound tree, which corre-
sponds to a progressively finer partition of F . In particular, the nodes of
this graph correspond to a collection F of subsets of F , which is such that:
1. F ∈ F (i.e., the set of all solutions is a node).
2. If x is a feasible solution, then {x} ∈ F (i.e., each solution viewed as
a singleton set is a node).
3. If a set Y ∈ F contains more than one solution x ∈ F , then there
exist disjoint sets Y1 , . . . , Yn ∈ F such that


n
Yi = Y.
i=1

The set Y is called the parent of Y1 , . . . , Yn , and the sets Y1 , . . . , Yn

are called the children or descendants of Y .
4. Each set in F other than F has a parent.
The collection of sets F defines the branch-and-bound tree as in Fig. 10.7.
In particular, this tree has the set of all feasible solutions F as its root node
and the singleton solutions {x}, x ∈ F , as terminal nodes. The arcs of the
graph are those that connect parents Y and their children Yi .
The key assumption in the branch-and-bound method is that for every
nonterminal node Y , there is an algorithm that calculates:
(a) A lower bound f Y to the minimum cost over Y

f Y ≤ min f (x).
x∈Y

(b) A feasible solution x ∈ Y , whose cost f (x) can serve as an upper

bound to the optimal cost of the original problem minx∈F f (x).
The main idea of the branch-and-bound algorithm is to save computation
by discarding the nodes/subsets of the tree that have no chance of con-
taining an optimal solution. In particular, the algorithm selects nodes Y
from the branch-and-bound tree, and checks whether the lower bound f Y
exceeds the best available upper bound [the minimal cost f (x) over all fea-
sible solutions x found so far]. If this is so, we know that Y cannot contain
an optimal solution, so all its descendant nodes in the tree need not be
considered further.
To organize the search through the tree, the algorithm maintains a
node list called OPEN, and also maintains a scalar called UPPER, which
is equal to the minimal cost over feasible solutions found so far. Initially,
OPEN contains just F , and UPPER is equal to ∞ or to the cost f (x) of
some feasible solution x ∈ F .
Sec. 10.2 Branch-and-Bound 485

F = {1,2,3,4,5}

Lower Bound _fY

_ Y = {1,2,3} {4,5}
Feasible Solution x ∈ Y

Y1 = {1,2} Y2 = {3} {4}

{4} {5}

{1} {2}

Figure 10.7: Illustration of a branch-and-bound tree. Each node Y (a subset

of the feasible set F ), except those consisting of a single solution, is partitioned
into several other nodes (subsets) Y1 , . . . , Yn . The original feasible set is divided
repeatedly into subsets until no more division is possible. For each node/subset
Y of the tree, one may compute a lower bound f to the optimal cost of the
Y
corresponding restricted subproblem minx∈Y f (x), and a feasible solution x ∈ Y ,
whose cost can serve as an upper bound to the optimal cost minx∈F f (x) of the
original problem. The idea is to use these bounds to economize computation by
eliminating nodes of the tree that cannot contain an optimal solution.

Branch-and-Bound Algorithm

Step 1: Remove a node Y from OPEN. For each child Yj of Y , do the

following: Find the lower bound f Y and a feasible solution x ∈ Yj . If
j

f Y < UPPER,
j

place Yj in OPEN. If in addition

f (x) < UPPER,

set
UPPER = f (x)
and mark x as the best solution found so far.
Step 2: (Termination Test) If OPEN is nonempty, go to step 1.
Otherwise, terminate; the best solution found so far is optimal.
486 Network Problems with Integer Constraints Chap. 10

A node Yj that is not placed in OPEN in Step 1 is said to be fath-

omed . Such a node cannot contain a better solution than the best solution
found so far, since the corresponding lower bound f Y is not smaller than
j
UPPER. Therefore nothing is lost when we drop this node from further
consideration and forego the examination of its descendants. Regardless of
how many nodes are fathomed, the branch-and-bound algorithm is guar-
anteed to examine either explicitly or implicitly (through fathoming) all
the terminal nodes, which are the singleton solutions. As a result, it will
terminate with an optimal solution.
Note that a small (near-optimal) value of UPPER and tight lower
bounds f Y contribute to the quick fathoming of large portions of the
j
branch-and-bound tree, and an early termination of the algorithm, with
either an optimal solution or a solution that is within some given tolerance
of being optimal. In fact, a popular variant, aimed at accelerating the
branch-and-bound algorithm, is to fix an
> 0, and replace the test

f Y j < UPPER

with
f Y j < UPPER −

in Step 1. This variant may terminate much faster, while the best solution
obtained upon termination is guaranteed to be within
of optimality.
Other variations of branch-and-bound relate to the method for se-
lecting a node from OPEN in Step 1. For example, a possible strategy
is to choose the node with minimal lower bound; alternatively, one may
choose the node containing the best solution found so far. In fact it is
neither practical nor necessary to generate a priori the branch-and-bound
tree. Instead, one may adaptively decide on the order and the manner in
which the nodes are partitioned into descendants based on the progress of
the algorithm.
Branch-and-bound typically uses “continuous” network optimization
problems (without integer constraints) to obtain lower bounds to the op-
timal costs of the restricted problems minx∈Y f (x) and to construct corre-
sponding feasible solutions. For example, suppose that our original problem
has a convex cost function, and a feasible set F that consists of convex set
constraints and side constraints, plus the additional constraint that all the
arc flows must be 0 or 1 . Then a restricted subset Y may specify that the
flows of some given subset of arcs are fixed at 0 or at 1, while the remaining
arc flows may take either the value 0 or the value 1. A lower bound to the
restricted optimal cost minx∈Y f (x) is then obtained by relaxing the 0-1
constraint on the latter arc flows, thereby allowing them to take any value
in the interval [0, 1] and resulting in a convex network problem with side
constraints. Thus the solution by branch-and-bound of a network problem
Sec. 10.2 Branch-and-Bound 487

with convex cost and side constraints plus additional integer constraints re-
quires the solution of many convex network problems with side constraints
but without integer constraints.

Example 10.8. Facility Location Problems

Let us consider the facility location problem introduced in Example 10.2,

which involves m clients and n locations. By xij = 1 (or xij = 0) we indicate
that client i is assigned to location j at a cost aij (or is not assigned, respec-
tively). We also introduce a 0-1 integer variable yj to indicate (with yj = 1)
that a facility is placed at location j at a cost bj . The problem is

 
n

minimize aij xij + b j yj

(i,j)∈A j=1

subject to xij = 1, i = 1, . . . , m,
{j|(i,j)∈A}

xij ≤ yj cj , j = 1, . . . , n,
{i|(i,j)∈A}

xij = 0 or 1, ∀ (i, j) ∈ A,
yj = 0 or 1, j = 1, . . . , n,

where cj is the maximum number of customers that can be served by a facility

at location j.
The solution of the problem by branch-and-bound involves the partition
of the feasible set F into subsets. The choice of subsets is somewhat arbitrary,
but it is convenient to select subsets of the form


F (J0 , J1 ) = (x, y) ∈ F | yj = 0, ∀ j ∈ J0 , yj = 1, ∀ j ∈ J1 ,

where J0 and J1 are disjoint subsets of the index set {1, . . . , n} of facility
locations. Thus, F (J0 , J1 ) is the subset of feasible solutions such that:
a facility is placed at the locations in J1 ,
no facility is placed at the locations in J0 ,
a facility may or may not be placed at the remaining locations.
For each node/subset F (J0 , J1 ), we may obtain a lower bound and a feasible
solution by solving the linear program where all integer constraints are relaxed
except for the variables yj , j ∈ J0 ∪ J1 , which have been fixed at either 0 or
1:

 
n

minimize aij xij + b j yj

(i,j)∈A j=1

subject to xij = 1, i = 1, . . . , m,
{j|(i,j)∈A}
488 Network Problems with Integer Constraints Chap. 10

xij ≤ yj cj , j = 1, . . . , n,
{i|(i,j)∈A}

xij ∈ [0, 1], ∀ (i, j) ∈ A,

yj ∈ [0, 1], ∀j∈
/ J0 ∪ J 1 ,
yj = 0, ∀ j ∈ J0 , yj = 1, ∀ j ∈ J1 .

As an illustration, let us work out the example shown in Figure 10.8,

which involves 3 clients and 2 locations. The facility capacities at the two
locations are c1 = c2 = 3. The cost coefficients aij and bj are shown next
to the corresponding arcs. The optimal solution corresponds to y1 = 0 and
y2 = 1, that is, placing a facility only in location 2 and serving all the clients
at that facility. The corresponding optimal cost is

f ∗ = 5.

Let us apply the branch-and-bound algorithm

 using the
 tree shown in
Fig. 10.8. We first consider the top node J0 = Ø, J1 = Ø , where neither
y1 nor y2 is fixed at 0 or 1. The lower bound f Y is obtained by solving the
(relaxed) linear program

minimize (2x11 + x12 ) + (2x21 + x22 ) + (x31 + 2x32 ) + 3y1 + y2

subject to x11 + x12 = 1, x21 + x22 = 1, x31 + x32 = 1,
x11 + x21 + x31 ≤ 3y1 , x12 + x22 + x32 ≤ 3y2 ,
0 ≤ xij ≤ 1, ∀ (i, j) ∈ A,
0 ≤ y1 ≤ 1, 0 ≤ y2 ≤ 1.

The optimal solution of this program is


1 if (i, j) = (1, 2), (2, 2), (3, 1),
xij =
0 otherwise,

y1 = 1/3, y2 = 2/3,

and the corresponding optimal cost (lower bound) is

f Y = 4.66.

A feasible solution of the original problem is obtained by rounding the frac-

tional values of y1 and y2 to

y 1 = 1, y 2 = 1,

and the associated cost is 7. Thus, we set

UPPER = 7,
Sec. 10.2 Branch-and-Bound 489

CLIENTS LOCATIONS
1
2
1 _
< 3y1
1 1
b1 = 3
1 2
2 1
1
_
< 3y2
2
1 2 b2 = 1
3

Feasible Solution Cost = 7

Lower Bound = 4.66

J0 = ∅, J1 = ∅
Feasible Solution Cost = 5 Lower Bound = 6.66
Lower Bound = 5 FATHOMED

J0 = {1}, J1 = ∅ J0 = ∅, J1 = {1}

J0 = {1}, J1 = {2} J0 = {1,2}, J1 = ∅ J0 = ∅, J1 = {1,2} J0 = {2}, J1 = {1}

Figure 10.8: Branch-and-bound solution of a facility location problem with

3 clients and 2 locations. The facility capacities at the two locations are c1 =
c2 = 3. The cost coefficients aij and bj are shown next to the corresponding
 
arcs. The relaxed problem for the top node J0 = Ø, J1 = Ø , corresponding
to relaxing all the integer constraints, is solved first, obtaining the lower and
upperbounds shown. Then  the relaxed problem corresponding to the left
node J0 = {1}, J1 = Ø is solved, obtaining the lower and upper bounds

shown. Finally, the relaxed problem corresponding to the right node J0 =

Ø, J1 = {1} is solved, obtaining a lower bound that is higher than the current
value of UPPER. As a result this node can be fathomed, and its descendants
need not be considered further.

  
and we place in OPEN the two descendants J0 = {1}, J1 = Ø and J0 =

Ø, J1 = {1} , corresponding to fixing y1 at 0 and at 1, respectively.
We proceed with
 the left branch of the branch-and-bound tree, and
consider the node J0 = {1}, J1 = Ø , corresponding to fixing y1 as well
as the corresponding flows x11 , x21 , and x31 to 0. The associated (relaxed)
linear program is

minimize x12 + x22 + 2x32 + y2

subject to x12 = 1, x22 = 1, x32 = 1,
x12 + x22 + x32 ≤ 3y2 ,
0 ≤ x12 ≤ 1, 0 ≤ x22 ≤ 1, 0 ≤ x32 ≤ 1,
0 ≤ y2 ≤ 1.
490 Network Problems with Integer Constraints Chap. 10

The optimal solution (in fact the only feasible solution) of this program is

1 if (i, j) = (1, 2), (2, 2), (3, 2),
xij =
0 otherwise,

y2 = 1,
and the corresponding optimal cost (lower bound) is

f Y = 5.

The optimal solution of the relaxed problem is integer, and its cost, 5, is lower
than the current value of UPPER, so we set

UPPER = 5.
   
The two descendants, J0 = {1}, J1 = {2} and J0 = {1, 2}, J1 = Ø ,
corresponding to fixing y2 at 1 and at 0, respectively, are placed in OPEN.
We proceed  with the right branch
 of the branch-and-bound tree, and
consider the node J0 = Ø, J1 = {1} , corresponding to fixing y1 to 1. The
associated (relaxed) linear program is

minimize (2x11 + x12 ) + (2x21 + x22 ) + (x31 + 2x32 ) + 3 + y2

subject to x11 + x12 = 1, x21 + x22 = 1, x31 + x32 = 1,
x11 + x21 + x31 ≤ 3, x12 + x22 + x32 ≤ 3y2 ,
0 ≤ xij ≤ 1, ∀ (i, j) ∈ A,
0 ≤ y2 ≤ 1.

The optimal solution of this program is


1 if (i, j) = (1, 2), (2, 2), (3, 1),
xij =
0 otherwise,

y2 = 2/3,
and the corresponding optimal cost (lower bound) is

f Y = 6.66.

This is larger than the current value of UPPER, so the node can be fathomed,
and its two descendants are not placed in OPEN. 
We conclude that one of the two descendants of the left node, J0 =
  
{1}, J1 = {2} and J0 = {1, 2}, J1 = Ø (the only nodes in OPEN), contains
the optimal solution. We can proceed to solve the relaxed linear programs
corresponding to these two nodes, and obtain the optimal solution. However,
there is also a shortcut here: since these are the only two remaining nodes
and the upper bound corresponding to these nodes coincides with the lower
bound, we can conclude that the lower bound is equal to the optimal cost
and the corresponding integer solution (y1 = 0, y2 = 1) is optimal.
Sec. 10.2 Branch-and-Bound 491

Generally, for the success of the branch-and-bound approach it is

important that the lower bounds are as tight as possible, because this
facilitates the fathoming of nodes, and leads to fewer restricted problem
solutions. On the other hand, the tightness of the bounds strongly depends
on how the problem is formulated as an integer programming problem.
There may be several possible formulations, some of which are “stronger”
than others in the sense that they provide better bounds within the branch-
and-bound context. As an illustration, consider the following example.

Example 10.9. Facility Location – Alternative Formulation

Consider the following alternative formulation of the preceding facility loca-

tion problem

 
n

minimize aij xij + b j yj

(i,j)∈A j=1

subject to xij = 1, i = 1, . . . , m,
{j|(i,j)∈A}

xij ≤ cj , j = 1, . . . , n,
{i|(i,j)∈A}

xij ≤ yj , ∀ (i, j) ∈ A,
xij = 0 or 1, ∀ (i, j) ∈ A,
yj = 0 or 1, j = 1, . . . , n.

This formulation involves a lot more constraints, but is in fact superior to

the one given earlier (cf. Example 10.8). The reason is that, if we relax the
0-1 constraints on xij and yj , the side
constraints x ≤ yj cj of Example
i ij
10.8 are implied by the constraints x ≤ cj and xij ≤ yj of the present
i ij
example. As a result, the lower bounds obtained by relaxing some of the
0-1 constraints are tighter in the alternative formulation just given, thereby
enhancing the effectiveness of the branch-and-bound method. In fact, it can
be verified that for the example of Fig. 10.8, by relaxing the 0-1 constraints
in the stronger formulation of the present example, we obtain the correct
optimal integer solution at the very first node of the branch-and-bound tree.

An important conclusion from the preceding example is that it is

possible to accelerate the branch-and-bound solution of a problem by adding
more side constraints. Even if these constraints do not affect the set of
feasible integer solutions, they can improve the lower bounds obtained by
relaxing the 0-1 constraints. Basically, when the integer constraints are
relaxed, one obtains a superset of the feasible set of integer solutions, so
with more side constraints, the corresponding superset becomes smaller
and approximates better the true feasible set (see Fig. 10.9). It is thus
very important to select a problem formulation such that when the integer
constraints are relaxed, the feasible set is as small as possible.
 492 Network Problems with Integer Constraints Chap. 10

Additional
Figure 10.9: Illustration of the effect
Side Constraints
of additional side constraints. They do
not affect the set of feasible integer solu-
tions, but they reduce the set of “relaxed
Set of Relaxed solutions,” that is, those x that satisfy
Integer all the constraints except for the inte-
Solutions
Solutions
ger constraints. This results in improved
lower bounds and a faster branch-and-
bound solution.

We note that the subject of characterizing the feasible set of an integer

programming problem, and approximating it tightly with a polyhedral set
has received extensive attention. In particular, there is a lot of theory
and accumulated practical knowledge on characterizing the feasible set in
specific problem contexts; see the references cited at the end of the chapter.
A further discussion of branch-and-bound is beyond our scope. We refer to
sources on linear and combinatorial optimization, such as Zoutendijk [1976],
Papadimitriou and Steiglitz [1982], Schrijver [1986], Nemhauser and Wolsey
[1988], Bertsimas and Tsitsiklis [1997], Cook, Cunningham, Pulleyblank,
and Schrijver [1998], which also describe many applications.

10.3 LAGRANGIAN RELAXATION

In this section, we consider an important approach for obtaining lower

bounds to use in the branch-and-bound method. Let us consider the case
of the network optimization problem with linear cost function, linear side
constraints, and integer constraints on the arc flows:

minimize a
x
 
subject to xij − xji = si , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

c
t x ≤ dt , t = 1, . . . , r,
xij ∈ Xij , ∀ (i, j) ∈ A,

where a and ct are given vectors, dt are given scalars, and each Xij is a
finite subset of contiguous integers (i.e., the convex hull of Xij contains
all the integers in Xij , as for example in the cases Xij = {0, 1} or Xij =
{1, 2, 3, 4}). We assume that the supplies si are integer , so that if the
side constraints c
t x ≤ dt were not present, the problem would become a
minimum cost flow problem that has integer optimal solutions, according
to the theory developed in Chapter 5. Note that for this it is not necessary
Sec. 10.3 Lagrangian Relaxation 493

that the arc cost coefficients aij (the components of the vectors a) be
integer.
In the Lagrangian relaxation approach, we eliminate the side con-
straints c
t x ≤ dt by adding to the cost function the terms µt (c
t x − dt ),
thereby forming the Lagrangian function

r
L(x, µ) = a
x + µt (c
t x − dt ),
t=1

where µ = (µ1 , . . . , µr ) is a vector of nonnegative scalars. Each µt may be

viewed as a penalty per unit violation of the corresponding side constraint
c
t x ≤ dt , and may also be viewed as a Lagrange multiplier.
A key idea of Lagrangian relaxation is that regardless of the choice
of µ, the minimization of the Lagrangian L(x, µ) over the set of remaining
constraints
F̃ = {x | xij ∈ Xij , x satisfies the conservation of flow constraints},
yields a lower bound to the optimal cost of the original problem (cf. the
weak duality property, discussed in Section 8.7). To see this, note that we
have
 
 r
min L(x, µ) = min a
x +

µt (ct x − dt )
x∈F̃ x∈F̃
t=1
 

r
≤ min a
x + µt (c
t x − dt )
x∈F̃ , ct x−dt ≤0, t=1,...,r
t=1
≤ min a
x,
x∈F̃ , ct x−dt ≤0, t=1,...,r

where the first inequality follows because the minimum of the Lagrangian in
the next-to-last expression is taken over a subset of F̃ , and the last inequal-
ity follows using the nonnegativity of µt . The lower bound minx∈F̃ L(x, µ)
can in turn be used in the branch-and-bound procedure discussed earlier.
Since in the context of branch-and-bound, it is important to use as
tight a lower bound as possible, we are motivated to search for an optimal
lower bound through adjustment of the vector µ. To this end, we form the
following dual function (cf. Section 8.7)
q(µ) = min L(x, µ),
x∈F̃

and we consider the dual problem

maximize q(µ)
subject to µt ≥ 0, t = 1, . . . , r.
Solution of this problem yields the tightest lower bound to the optimal cost
of the original problem.
494 Network Problems with Integer Constraints Chap. 10

Example 10.10. Constrained Shortest Path Problem

As an example of the use of Lagrangian relaxation, consider the constrained

shortest path problem discussed in Example 8.6 of Section 8.4. Here, we want
to find a simple forward path P from an origin node s to a destination node
t that minimizes the path length

aij ,
(i,j)∈P

subject to the following side constraints on P :


ckij ≤ dk , k = 1, . . . , K.
(i,j)∈P

As discussed in Section 8.4, we can formulate this as the following network

flow problem with integer constraints and side constraints:

minimize aij xij
(i,j)∈A

  1 if i = s,
subject to xij − xji = −1 if i = t,
{j|(i,j)∈A} {j|(j,i)∈A} 0 otherwise, (10.7)

xij = 0 or 1, ∀ (i, j) ∈ A,

ckij xij ≤ dk , k = 1, . . . , K.
(i,j)∈A

Here, a path P from s to t is optimal if and only if the flow vector x defined
by 
1 if (i, j) belongs to P ,
xij =
0 otherwise,
is an optimal solution of the problem (10.7).
To apply Lagrangian relaxation, we eliminate the side constraints, and
we form the corresponding Lagrangian function assigning a nonnegative mul-
tiplier µk to the kth constraint. Minimization of the Lagrangian now becomes
a shortest path problem with respect to corrected arc lengths âij given by


K

âij = aij + µk ckij .

k=1

(We assume here that there are no negative length cycles with respect to the
arc lengths âij ; this will be so if all the aij and ckij are nonnegative.) We
then obtain µ∗ that solves the dual problem maxµ≥0 q(µ) and we obtain a
corresponding optimal cost/lower bound. We can then use µ∗ to obtain a
feasible solution (a path that satisfies the side constraints) as discussed in
Example 8.6.
Sec. 10.3 Lagrangian Relaxation 495

The preceding example illustrates an important advantage of La-

grangian relaxation, as applied to integer-constrained network problems: it
eliminates the side constraints simultaneously with the integer constraints.
In particular, minimizing L(x, µ) over the set

F̃ = {x | xij ∈ Xij , x satisfies the conservation of flow constraints}

is a (linear) minimum cost flow problem that can be solved using the
methodology of Chapters 2-7: the Lagrangian L(x, µ) is linear in x and
the integer constraints do not matter, and can be replaced by the inter-
val constraints xij ∈ X̂ij , where X̂ij is the convex hull of the set Xij .
This should be contrasted with the integer constraint relaxation approach,
where we eliminate just the integer constraints, while leaving the side con-
straints unaffected (see the facility location problem that we solved using
branch-and-bound in Example 10.8). As a result, the minimum cost flow
methodology of Chapters 2-7 does not apply when there are side constraints
and the integer constraint relaxation approach is used. This is the main
reason for the widespread use of Lagrangian relaxation in combination with
branch-and-bound.
Actually, in Lagrangian relaxation it is not mandatory to eliminate
just the side constraints. One may eliminate the conservation of flow con-
straints, in addition to or in place of the side constraints. (The multipliers
corresponding to the conservation of flow constraints should be uncon-
strained in the dual problem, because the conservation of flow is expressed
in terms of equality constraints; cf. the discussion in Section 8.7.) One
still obtains a lower bound to the optimal cost of the original problem,
because of the weak duality property (cf. Section 8.7). However, the mini-
mization of the Lagrangian is not a minimum cost flow problem anymore.
Nonetheless, by choosing properly the constraints to eliminate and by tak-
ing advantage of the special structure of the problem, the minimization
of the Lagrangian over the remaining set of constraints may be relatively
simple. The following is an illustrative example.

Example 10.11. Traveling Salesman Problem

Consider the traveling salesman problem of Example 10.1. Here, we want to

find a minimum cost tour in a complete graph where the cost of arc (i, j) is
denoted aij . We formulate this as the following network problem with side
constraints and 0-1 integer constraints:

minimize aij xij
(i,j)∈A

subject to xij = 1, i = 1, . . . , N, (10.8)
j=1,...,N
j=i
496 Network Problems with Integer Constraints Chap. 10

xij = 1, j = 1, . . . , N, (10.9)
i=1,...,N
i=j

xij = 0 or 1, ∀ (i, j) ∈ A, (10.10)

 
the subgraph with node-arc set N , {(i, j) | xij = 1} is connected. (10.11)
We may express the connectivity constraint (10.11) in several different
ways, leading to different Lagrangian relaxation and branch-and-bound algo-
rithms. One of the most successful formulations is based on the notion of
a 1-tree, which consists of a tree that spans nodes 2, . . . , N , plus two arcs
that are incident to node 1. Equivalently, a 1-tree is a connected subgraph
that contains a single cycle passing through node 1 (see Fig. 10.10). Note
that if the conservation of flow constraints (10.8) and (10.9), and the integer
constraints (10.10) are satisfied, then the connectivity
 constraint (10.11)
 is
equivalent to the constraint that the subgraph N , {(i, j) | xij = 1} is a
1-tree.

3
1
Figure 10.10: Illustration of a 1-tree. It con-
sists of a tree that spans nodes 2, . . . , N , plus
two arcs that are incident to node 1.

Let X1 be the set of all x with 0 − 1 components, and such that the
subgraph N , {(i, j) | xij = 1} is a 1-tree. Let us consider a Lagrangian
relaxation approach based on elimination of the conservation of flow equa-
tions. Assigning multipliers ui and vj to the constraints (10.8) and (10.9),
respectively, the Lagrangian function is

 
N

N

L(x, u, v) = (aij + ui + vj )xij − ui − vj .

i,j,i=j i=1 j=1

The minimization of the Lagrangian is over all 1-trees, leading to the problem
 

min (aij + ui + vj )xij .
x∈X1
i,j,i=j

If we view aij + ui + vj as a modified cost of arc (i, j), this minimization

is quite easy. It is equivalent to obtaining a tree of minimum modified cost
Sec. 10.3 Lagrangian Relaxation 497

that spans the nodes 2, . . . , N , and then adding two arcs that are incident to
node 1 and have minimum modified cost. The minimum cost spanning tree
problem can be easily solved using the Prim-Dijkstra algorithm (see Exercise
2.30).

Unfortunately, the Lagrangian relaxation method has several weak-

nesses:
(a) Even if we find an optimal µ, we still have only a lower bound to the
optimal cost of the original problem.
(b) The minimization of L(x, µ) over the set
F̃ = {x | xij ∈ Xij , x satisfies the conservation of flow constraints},
may yield an x that violates some of the side constraints c
t x − dt ≤ 0,
so it may be necessary to adjust this x for feasibility using some
heuristic.
(c) The maximization of q(µ) over µ ≥ 0 may be quite nontrivial for a
number of reasons, including the fact that q is typically nondifferen-
tiable.
In what follows in this section, we will discuss the algorithmic method-
ology for solving the dual problem, including the subgradient and cutting
plane methods, which have enjoyed a great deal of popularity. These meth-
ods have also been used widely in connection with various decomposition
schemes for large-scale problems with special structure. For further dis-
cussion, we refer to the nonlinear programming literature (see for example
Lasdon [1970], Auslender [1976], Shapiro [1979], Shor [1985], Poljak [1987],
Hiriart-Urruty and Lemarechal [1993], and Bertsekas [1995b]).

10.3.1 Subgradients of the Dual Function

Let us consider the algorithmic solution of the dual problem

maximize q(µ)
subject to µt ≥ 0, t = 1, . . . , r.
The dual function is
q(µ) = min L(x, µ),
x∈F̃

where
F̃ = {x | xij ∈ Xij , x satisfies the conservation of flow constraints},
and L(x, µ) is the Lagrangian function

r
L(x, µ) = a
x + µt (c
t x − dt ).
t=1
498 Network Problems with Integer Constraints Chap. 10

Recall here that the set F̃ is finite, because we have assumed that each Xij
is a finite set of contiguous integers.
We note that for a fixed x ∈ F̃ , the Lagrangian L(x, µ) is a linear
function of µ. Thus, because the set F̃ is finite, the dual function q is the
minimum of a finite number of linear functions of µ – there is one such
function for each x ∈ F̃ . For conceptual simplification, we may write q in
the following generic form:

q(µ) = min{αi
µ + βi }, (10.12)

i∈I

where I is some finite index set, and αi and βi are suitable vectors and
scalars, respectively (see Fig. 10.11).
Of particular interest for our purposes are the “slopes” of q at vari-
ous vectors µ, i.e., the vectors αiµ , where iµ ∈ I is an index attaining the
minimum of αi
µ + βi over i ∈ I [cf. Eq. (10.12)]. If iµ is the unique index
attaining the minimum, then q is differentiable (in fact linear) at µ, and its
gradient is aiµ . If there are multiple indices i attaining the minimum, then
q is nondifferentiable at µ (see Fig. 10.11). To deal with such differentia-
bilities, we generalize the notion of a gradient. In particular, we define a
subgradient of q at a given µ ≥ 0 to be any vector g such that

q(ν) ≤ q(µ) + (ν − µ)
g, ∀ ν ≥ 0, (10.13)

(see Fig. 10.11). The right-hand side of the above inequality provides a
linear approximation to the dual function q using the function value q(µ)
at the given µ and the corresponding subgradient g. The approximation
is exact at the vector µ, and is an overestimate at other vectors ν. Some
further properties of subgradients are summarized in Appendix A.
We now consider the calculation of subgradients of the dual function.
For any µ, let xµ minimize the Lagrangian L(x, µ) over x ∈ F̃ ,

xµ = arg min L(x, µ).

x∈F̃

Let us show that the vector g(xµ ) that has components

gt (xµ ) = c
t xµ − dt , t = 1, . . . , r,

is a subgradient of q at µ. To see this, we use the definition of L, q, and

xµ to write for all ν ≥ 0,

q(ν) = min L(x, ν)

x∈F̃

≤ L(xµ , ν)
= a
xµ + ν
g(xµ )
= a
xµ + µ
g(xµ ) + (ν − µ)
g(xµ )
= q(µ) + (ν − µ)
g(xµ ),
Sec. 10.3 Lagrangian Relaxation 499

q(µ) = min { α i' µ + βi }

i ∈I

α 2 ' µ + β2 α 3 ' µ + β3

α 1 ' µ + β1

_
µ µ~ µ

Figure 10.11: Illustration of the dual function q and its subgradients. The
generic form of q is
q(µ) = min{αi µ + βi },
i∈I

where I is some finite index set, and αi and βi are suitable vectors and scalars,
respectively. Given µ, and an index iµ ∈ I attaining the minimum in the above
equation, the vector αiµ is a subgradient at µ. Furthermore, any subgradient
at µ is a convex combination of vectors aiµ such that iµ ∈ I and iµ attains the
minimum of αi µ + βi over i ∈ I. For example, at the vector µ shown in the figure,
there is a unique subgradient, the vector α1 . At the vector µ̃ shown in the figure,
the set of subgradients is the line segment connecting the vectors α2 and α3 .

so the subgradient inequality (10.13) is satisfied for g = g(xµ ). Thus,

for a given µ, the evaluation of q(µ), which requires finding a minimizer
xµ of L(x, µ) over F̃ , yields as a byproduct the subgradient g(xµ ). This
convenience in calculating subgradients is particularly important for the
algorithms that we discuss in what follows in this section.

10.3.2 Subgradient Methods

We now turn to algorithms that use subgradients for solving the dual prob-
lem. The subgradient method consists of the iteration
! "+
µk+1 = µk + sk g k , (10.14)

where g k is any subgradient of q at µk , sk is a positive scalar stepsize,

and [y]+ is the operation that sets to 0 all the negative components of the
vector y. Thus the iteration (10.14) can be written as


µk+1
t = max 0, µkt + sk gtk , t = 1, . . . , r.
500 Network Problems with Integer Constraints Chap. 10

The simplest way to calculate the subgradient g k is to find an xµk that

minimizes L(x, µk ) over x ∈ F̃ , and to set

g k = g(xµk ),

where for every x, g(x) is the r-dimensional vector with components

gt (x) = c
t x − dt , t = 1, . . . , r.

An important fact about the subgradient method is that the new

iterate may not improve the dual cost for all values of the stepsize sk ; that
is, we may have

 
q [µk + sk g k ]+ < q(µk ), ∀ sk > 0

(see Fig. 10.12). What makes the subgradient method work is that for suf-
ficiently small stepsize sk , the distance of the current iterate to the optimal
solution set is reduced , as illustrated in Fig. 10.12, and as shown in the
following proposition.

µ2

µk+1 = [ µk + sk g k ]+

Contours of q
µk
o
< 90
µ*
µ1

µk + sk g k

Figure 10.12: Illustration of how it may not be possible to improve the dual
function by using the subgradient iteration µk+1 = [µk + sk g k ]+ , regardless of
the value of the stepsize sk . However, the distance to any optimal solution µ∗ is
reduced using a subgradient iteration with a sufficiently small stepsize. The crucial
fact, which follows from the definition of a subgradient, is that the angle between
the subgradient g k and the vector µ∗ − µk is less than 90 degrees. As a result,
for sk small enough, the vector µk + sk g k is closer to µ∗ than µk . Furthermore,
the vector [µk + sk g k ]+ is closer to µ∗ than µk + sk g k is.
Sec. 10.3 Lagrangian Relaxation 501

Proposition 10.1: If µk is not optimal, then for any dual optimal

solution µ∗ , we have

µk+1 − µ∗  < µk − µ∗ ,

for all stepsizes sk such that

 
2 q(µ∗ ) − q(µk )
0< sk < . (10.15)
g k 2

Proof: We have

µk + sk g k − µ∗ 2 = µk − µ∗ 2 − 2sk (µ∗ − µk )
g k + (sk )2 g k 2 ,

and by using the subgradient inequality (10.13),

(µ∗ − µk )
g k ≥ q(µ∗ ) − q(µk ).

Combining the last two relations, we obtain

 
µk + sk g k − µ∗ 2 ≤ µk − µ∗ 2 − 2sk q(µ∗ ) − q(µk ) + (sk )2 g k 2 .

We can now verify that for the range of stepsizes of Eq. (10.15) the sum of
the last two terms in the above relation is negative. In particular, with a
straightforward calculation, we can write this relation as
 2
γ k (2 − γ k ) q(µ∗ ) − q(µk )
µk + sk g k − µ∗ 2 ≤ µk − µ∗ 2 − , (10.16)
g k 2

where
sk g k 2
γk = .
q(µ∗ ) − q(µk )
If the stepsize sk satisfies Eq. (10.15), then 0 < γ k < 2, so Eq. (10.16)
yields
µk + sk g k − µ∗  < µk − µ∗ .
We now observe that since µ∗ ≥ 0, we have
#! " #
# µk + sk g k + − µ∗ # ≤ µk + sk g k − µ∗ ,

and from the last two inequalities, we obtain µk+1 − µ∗  < µk − µ∗ .
Q.E.D.
502 Network Problems with Integer Constraints Chap. 10

The inequality (10.16) can also be used to establish convergence and

rate of convergence results for the subgradient method with stepsize rules
satisfying  
k
2 q(µ∗ ) − q(µk )
0<s <
g k 2
[cf. Eq. (10.15)]. Unfortunately, however, unless we know the dual optimal
cost q(µ∗ ), which is rare, the range of stepsizes (10.15) is unknown. In
practice, a frequently used stepsize formula is
 
α k q k − q(µk )
sk = , (10.17)
g k 2

where q k is an approximation to the optimal dual cost and

0 < αk < 2.

Note that we can estimate the optimal dual cost from below with the
best current dual cost
q̂ k = max q(µi ).
0≤i≤k

As an overestimate of the optimal dual cost, we can use the cost f (x̄) of any
primal feasible solution x̄; in many circumstances, primal feasible solutions
are naturally obtained in the course of the algorithm. Finally, the special
structure of many problems can be exploited to yield improved bounds to
the optimal dual cost.
Here are two common ways to choose αk and q k in the stepsize formula
(10.17):
(a) q k is the best known upper bound to the optimal dual cost at the kth
iteration and αk is a number, which is initially equal to one and is
decreased by a certain factor (say, two) every few (say, five or ten)
iterations. An alternative formula for αk is
m
αk = ,
k+m
where m is a positive integer.
(b) αk = 1 for all k and q k is given by
 
q k = 1 + β(k) q̂ k , (10.18)

where q̂ k is the best current dual cost q̂ k = max0≤i≤k q(µi ). Further-

more, β(k) is a number greater than zero, which is increased by a
certain factor if the previous iteration was a “success,” that is, if it
improved the best current dual cost, and is decreased by some other
factor otherwise. This method requires that q̂ k > 0. Also, if upper
Sec. 10.3 Lagrangian Relaxation 503

bounds q̃ k to the optimal dual cost are available as discussed earlier,

then a natural improvement to Eq. (10.18) is

  
q k = min q̃ k , 1 + β(k) q̂ k .

For a convergence analysis of the subgradient method and its variants,

we refer to the literature cited at the end of the chapter (see also Exercises
10.36-10.38). However, the convergence properties of the schemes most
often preferred in practice, including the ones given above, are neither
solid nor well understood. It is easy to find problems where the subgradient
method works very poorly. On the other hand, the method is simple and
works well for many types of problems, yielding good approximate solutions
within a few tens or hundreds of iterations. Also, frequently a good primal
feasible solution can be obtained using effective heuristics, even with a
fairly poor dual solution.

10.3.3 Cutting Plane Methods

Consider again the dual problem

maximize q(µ)
subject to µ ≥ 0.

The cutting plane method, at the kth iteration, replaces the dual function
q by a polyhedral approximation Qk , constructed using the vectors µi and
corresponding subgradients g i , i = 0, 1, . . . , k − 1, obtained so far. It then
solves the problem
maximize Qk (µ)
subject to µ ≥ 0.
In particular, for k = 1, 2, . . ., Qk is given by


Qk (µ) = min q(µi ) + (µ − µi )
g i , (10.19)
i=0,...,k−1

and the kth iterate is generated by

µk = arg max Qk (µ). (10.20)

µ≥0

As in the case of subgradient methods, the simplest way to calculate

the subgradient g i is to find an xµi that minimizes L(x, µi ) over x ∈ F̃ ,
and to set
g i = g(xiµ ),
where for every x, g(x) is the r-dimensional vector with components

gt (x) = c
t x − dt , t = 1, . . . , r.
504 Network Problems with Integer Constraints Chap. 10

Note that the approximation Qk (µ) is an overestimate of the dual function

q,
q(µ) ≤ Qk (µ), µ ≥ 0, (10.21)
since, in view of the definition of a subgradient [cf. Eq. (10.13)], each of
the linear terms in the right-hand side of Eq. (10.19) is an overestimate of
q(µ).
We assume that, for all k, it is possible to find a maximum µk of
Q over µ ≥ 0. To ensure this, the method has to be suitably initialized;
k

for example by selecting a sufficiently large number of vectors µ, and by

computing corresponding subgradients, to form an initial approximation
that is bounded from above over the set {µ | µ ≥ 0}. Thus, in this variant,
we start the method at some iteration k > 0, with the vectors µ0 , . . . , µk−1
suitably selected so that Qk (µ) is bounded from above over µ ≥ 0. Al-
ternatively, we may maximize Qk over a suitable bounded polyhedral set
that is known to contain an optimal dual solution, instead of maximizing
over µ ≥ 0. We note that given the iterate µk , the method produces both
the exact and the approximate dual values q(µk ) and Qk (µk ). It can be
seen, using Eqs. (10.20) and (10.21), that the optimal dual cost is bracketed
between these two values:
q(µk ) ≤ max q(µ) ≤ Qk (µk ). (10.22)
µ≥0

Thus, in particular, the equality

q(µk ) = Qk (µk ) (10.23)
guarantees the optimality of the vector µk . It turns out that because the
dual function is piecewise linear, and consequently only a finite number of
subgradients can be generated, the optimality criterion (10.23) is satisfied
in a finite number of iterations, and the method terminates. This is shown
in the following proposition and is illustrated in Fig. 10.13.

Proposition 10.2: The cutting plane method terminates finitely;

that is, for some k, µk is a dual optimal solution and the termina-
tion criterion (10.23) is satisfied.

Proof: For notational convenience, let us write the dual function in the
polyhedral form
q(µ) = min{αi
µ + βi },
i∈I

where I is some finite index set and αi , βi , i ∈ I, are suitable vectors

and scalars, respectively. Let ik be an index attaining the minimum in the
equation
q(µk ) = min{αi
µk + βi },
i∈I
Sec. 10.3 Lagrangian Relaxation 505

q(µ0) + (µ − µ0)'g 0

q(µ1) + (µ − µ1)'g 1

q(µ)

µ0 µ3 µ2 µ1 µ
µ4 = µ*

Figure 10.13: Illustration of the cutting plane method. With each new iterate
µi , a new hyperplane q(µi ) + (µ − µi ) g i is added to the polyhedral approximation
of the dual function. The method converges finitely, since if µk is not optimal, a
new cutting plane will be added at the corresponding iteration, and there can be
only a finite number of cutting planes.

so that αik is a subgradient at µk . If the termination criterion (10.23) is

not satisfied at µk , we must have

αi
k µk + βik = q(µk ) < Qk (µk ).

Since
Qk (µk ) = min {αi
m µk + βim },
0≤m≤k−1

it follows that the pair (αik , βik ) is not equal to any of the preceding pairs
(αi0 , βi0 ), . . . , (αik−1 , βik−1 ). Since the index set I is finite, it follows that
there can be only a finite number of iterations for which the termination
criterion (10.23) is not satisfied. Q.E.D.

Despite its finite convergence property, the cutting plane method may
converge slowly, and in practice one may have to stop it short of finding an
optimal solution [the error bounds (10.22) may be used for this purpose].
An additional drawback of the method is that it can take large steps away
from the optimum even when it is close to (or even at) the optimum. This
phenomenon is referred to as instability, and has another undesirable effect,
namely, that µk−1 may not be a good starting point for the algorithm that
minimizes Qk (µ). A way to limit the effects of this phenomenon is to add
to the polyhedral function approximation a quadratic term that penalizes
large deviations from the current point. In this method, µk is obtained as
 
1
µ = arg max Q (µ) − k µ − µ
k k k−1  ,
2
µ≥0 2c
506 Network Problems with Integer Constraints Chap. 10

where {ck } is a positive nondecreasing scalar parameter sequence. This

is known as the proximal cutting plane algorithm, and is related to the
proximal minimization method discussed in Section 8.8.5. It can be shown
that this variant of the cutting plane method also terminates finitely thanks
to the polyhedral nature of q.
Another interesting variant of the cutting plane method, known as
the central cutting plane method , maintains the polyhedral approximation
Qk (µ) to the dual function q, but generates the next vector µk by using a
somewhat different mechanism. In particular, instead of maximizing Qk ,
the method obtains µk by finding a “central pair” (µk , z k ) within the subset


S k = (µ, z) | µ ≥ 0, q̂ k ≤ q(µ), q̂ k ≤ z ≤ Qk (µ) ,
where q̂ k is the best lower bound to the optimal dual cost that has been
found so far,
q̂ k = max q(µi ).
i=0,...,k−1

The set Sk is illustrated in Fig. 10.14.

q(µ0) + (µ − µ0)'g 0

Set S 2
Central Pair ( µ2,z 2 )
q(µ1) + (µ − µ1)'g 1

q2

q(µ)

µ0 µ* µ2 µ1 µ

Figure 10.14: Illustration of the set


Sk = (µ, z) | µ ≥ 0, q̂ k ≤ q(µ), q̂ k ≤ z ≤ Qk (µ) ,

for k = 2, in the central cutting plane method.

There are several possible methods for finding the central pair (µk , z k ).
Roughly, the idea is that the central pair should be “somewhere in the
middle” of S k . For example, consider the case where S k is polyhedral with
nonempty interior. Then (µk , z k ) could be the analytic center of S k , where
for any polyhedron
P = {y | a
p y ≤ cp , p = 1, . . . , m}
Sec. 10.3 Lagrangian Relaxation 507

with
m nonempty
interior, its analytic center is the unique maximizer of
p=1 ln(cp − ap y) over y ∈ P . Another possibility is the ball center of
S, that is, the center of the largest inscribed sphere in S. Assuming that
the polyhedron P given above has nonempty interior, its ball center can
be obtained by solving the following problem with optimization variables
(y, σ):

maximize σ
subject to a
p (y + d) ≤ cp , ∀ d ≤ σ, p = 1, . . . , m.

It can be seen that this problem is equivalent to the linear program

maximize σ
subject to a
p y + ap σ ≤ cp , p = 1, . . . , m.

While the central cutting plane methods are not guaranteed to ter-
minate finitely, their convergence properties are satisfactory. Furthermore,
the methods have benefited from advances in the implementation of interior
point methods; see the references cited at the end of the chapter.

10.3.4 Decomposition and Multicommodity Flows

Lagrangian relaxation is particularly convenient when by eliminating the

side constraints, we obtain a network problem that decomposes into several
independent subproblems. A typical example arises in multicommodity
flow problems where we want to minimize


M 
aij (m)xij (m) (10.24)
m=1 (i,j)∈A

subject to the conservation of flow constraints

 
xij (m) − xji (m) = si (m), ∀ i ∈ N , m = 1, . . . , M,
{j|(i,j)∈A} {j|(j,i)∈A}
(10.25)
the set constraints

xij (m) ∈ Xij (m), ∀ m = 1, . . . , M, (i, j) ∈ A, (10.26)

and the side constraints


M
A(m)x(m) ≤ b. (10.27)
m=1
508 Network Problems with Integer Constraints Chap. 10

Here si (m) are given supply integers for the mth commodity, A(m) are
given matrices, b is a given vector, and x(m) is the flow vector of the
mth commodity, with components xij (m), (i, j) ∈ A. Furthermore, each
Xij (m) is a finite subset of contiguous integers.
The dual function is obtained by relaxing the side constraints (10.27),
and by minimizing the corresponding Lagrangian function. This minimiza-
tion separates into m independent minimizations, one per commodity:


M
 

q(µ) = −µ
b + min a(m) + A(m)
µ x(m), (10.28)
x(m)∈F (m)
m=1

where a(m) is the vector with components aij (m), (i, j) ∈ A, and


F (m) = x(m) satisfying Eq. (10.25) | xij (m) ∈ Xij (m), ∀ (i, j) ∈ A .

An important observation here is that each of the minimization sub-

problems above is a minimum cost flow problem that can be solved using
the methods of Chapters 2-7. Furthermore, if xµ (m) solves the mth sub-
problem, the vector
M
gµ = A(m)xµ (m) − b (10.29)
m=1

is a subgradient of q at µ.
Let us now discuss the computational solution of the dual problem
maxµ≥0 q(µ). The application of the subgradient method is straightfor-
ward, so we concentrate on the cutting plane method, which leads to a
method known as Dantzig-Wolfe decomposition. This method consists of
the iteration
µk = arg max Qk (µ),
µ≥0

where Qk (µ)is the piecewise linear approximation of the dual function

based on the preceding function values q(µ0 ), . . . , q(µk−1 ), and the corre-
sponding subgradients g 0 , . . . , g k−1 :


Qk (µ) = min q(µ0 ) + (µ − µ0 )
g 0 , . . . , q(µk−1 ) + (µ − µk−1 )
g k−1 .

Consider now the cutting plane subproblem maxµ≥0 Qk (µ). By in-

troducing an auxiliary variable v, we can write this problem as

maximize v
subject to q(µi ) + (µ − µi )
g i ≥ v, i = 0, . . . , k − 1, (10.30)
µ ≥ 0.

This is a linear program in the variables v and µ. We can form its dual
problem by assigning a Lagrange multiplier ξ i to each of the constraints
Sec. 10.3 Lagrangian Relaxation 509

q(µi ) + (µ − µi )
g i ≥ v. After some calculation, this dual problem can be

verified to have the form


k−1
 
minimize ξ i q(µi ) − µi
g i
i=0

k−1 
k−1 (10.31)
subject to ξ i = 1, ξ i g i ≤ 0,
i=0 i=0
ξ i ≥ 0, i = 0, . . . , k − 1.

Using Eqs. (10.28) and (10.29), we have


M
 
i
q(µi ) = −µi
b + a(m) + A(m)
µi xµ (m),
m=1


M
i
gi = A(m)xµ (m) − b,
m=1

so the problem (10.31) can be written as


M 
k−1
i
minimize a(m)
ξ i xµ (m)
m=1 i=0

k−1 
M 
k−1
i
(10.32)
subject to ξ i = 1, A(m) ξ i xµ (m) ≤ b,
i=0 m=1 i=0
ξi ≥ 0, i = 0, . . . , k − 1.

The preceding problem is called the master problem. It is the dual of

the cutting plane subproblem maxµ≥0 Qk (µ), which in turn approximates
the dual problem maxµ≥0 q(µ); in short, it is the dual of the approximate
dual . We may view this problem as an approximate version of the primal
problem where the commodity flow vectors x(m) are constrained to lie in
i
the convex hull of the already generated vectors xµ (m), i = 0, . . . , k − 1,
rather than in their original constraint set. It can be shown, using linear
programming theory, that if the problem (10.30) has an optimal solution
[i.e., enough vectors µi are available so that the maximum of Qk (µ) over
µ ≥ 0 is attained], then the master problem also has an optimal solution.
Suppose now that we solve the master problem (10.32) using a method
that yields a Lagrange multiplier vector, call it µk , corresponding to the
constraints

M 
k−1
i
A(m) ξ i xµ (m) ≤ b.
m=1 i=0
510 Network Problems with Integer Constraints Chap. 10

(Standard linear programming methods, such as the simplex method, can

be used for this.) Then, the dual of the master problem [which is the cutting
plane subproblem maxµ≥0 Qk (µ)] is solved by the Lagrange multiplier µk .
Therefore, µk is the next iterate of the cutting plane method.
We can now piece together the typical cutting plane iteration.

Cutting Plane – Dantzig-Wolfe Decomposition Iteration

i
Step 1: Given µ0 , . . . , µk−1 , and the commodity flow vectors xµ (m)
for m = 1, . . . , M and i = 0, . . . , k − 1, solve the master problem


M 
k−1
i
minimize a(m)
ξ i xµ (m)
m=1 i=0

k−1 
M 
k−1
i
subject to ξ i = 1, A(m) ξ i xµ (m) ≤ b,
i=0 m=1 i=0
ξ i ≥ 0, i = 0, . . . , k − 1.

and obtain µk , which is a Lagrange multiplier vector of the constraints


M 
k−1
i
A(m) ξ i xµ (m) ≤ b.
m=1 i=0

k
Step 2: For each m = 1, . . . , M , obtain a solution xµ (m) of the
minimum cost flow problem
 

min a(m) + A(m)
µk x(m).
x(m)∈F (m)

k
Step 3: Use xµ (m) to modify the master problem by adding one
more variable ξ k and go to the next iteration.

Decomposition by Right-Hand Side Allocation

There is an alternative decomposition approach for solving the multicom-

modity flow problem with side constraints (10.24)-(10.27). In this ap-
proach, we introduce auxiliary variables y(m), m = 1, . . . , M , and we write
Sec. 10.3 Lagrangian Relaxation 511

the problem as


M
minimize a(m)
x(m)
m=1
subject to x(m) ∈ F (m), m = 1, . . . , M,

M
y(m) = b, A(m)x(m) ≤ y(m), m = 1, . . . , M.
m=1

Equivalently, we can write the problem as


M
minimize min a(m)
x(m)
x(m)∈F (m)
m=1 A(m)x(m)=y(m)
(10.33)

M
subject to y(m) = b, y(m) ∈ Y (m), m = 1, . . . , M,
m=1

where Y (m) is the set of all vectors y(m) for which the inner minimization
problem

minimize a(m)
x(m)

(10.34)
subject to x(m) ∈ F (m), A(m)x(m) ≤ y(m)

has at least one feasible solution.

Let us define
 
pm y(m) = min a(m)
x(m).
x(m)∈F (m)
A(m)x(m)≤y(m)

Then, problem (10.33) can be written as


M
 
minimize pm y(m)
m=1

M
subject to y(m) = b, y(m) ∈ Y (m), m = 1, . . . , M.
m=1

This problem, called the master problem, may be solved with nondifferen-
tiable optimization methods, and in particular with the subgradient and
the cutting plane methods. Note, however, that the commodity problems
(10.34) involve the side constraints A(m)x(m) ≤ y(m), and need not be
of the minimum cost flow type, except in special cases. We refer to the
literature cited at the end of the chapter for further details.
 512 Network Problems with Integer Constraints Chap. 10

10.4 LOCAL SEARCH METHODS

Local search methods are a broad and important class of heuristics for
discrete optimization. They apply to the general problem of minimizing
a function f (x) over a finite set F of (feasible) solutions. In principle,
one may solve the problem by global enumeration of the entire set F of
solutions (this is what branch-and-bound does). A local search method
tries to economize on computation by using local enumeration, based on
the notion of a neighborhood N (x) of a solution x, which is a (usually very
small) subset of F , containing solutions that are “close” to x in some sense.
In particular, given a solution x, the method selects among the solu-
tions in the neighborhood N (x) a successor solution x, according to some
rule. The process is then repeated with x replacing x (or stops when some
termination criterion is met). Thus a local search method is characterized
by:
(a) The method for choosing a starting solution.
(b) The definition of the neighborhood N (x) of a solution x.
(c) The rule for selecting a successor solution from within N (x).
(d) The termination criterion.
For an example of a local search method, consider the k-OPT heuristic
for the traveling salesman problem that we discussed in Example 10.1. Here
the starting tour is obtained by using some method, based for example on
subtour elimination or a minimum weight spanning tree, as discussed in
Example 10.1. The neighborhood of a tour T is defined as the set N (T )
of all tours obtained from T by exchanging k arcs that belong to T with
another k arcs that do not belong to T . The rule for selecting a successor
tour is based on cost improvement; that is, the tour selected from N (T )
has minimum cost over all tours in N (T ) that have smaller cost than T .
Finally, the algorithm terminates when no tour in N (T ) has smaller cost
than T . Another example of a local search method is provided by the
Esau-Williams heuristic of Fig. 10.5.
The definition of a neighborhood often involves intricate calculations
and suboptimizations that aim to bring to consideration promising neigh-
bors. Here is an example, due to Kernighan and Lin [1970]:

Example 10.12. (Uniform Graph Partitioning)

Consider a graph (N , A) with 2n nodes, and a cost aij for each arc (i, j). We
want to find a partition of N into two subsets N1 and N2 , each with n nodes,
so that the total cost of the arcs connecting N1 and N2 ,
 
aij + aij ,
(i,j), i∈N1 , j∈N2 (i,j), i∈N2 , j∈N1
Sec. 10.4 Local Search Methods 513

is minimized.
Here a natural neighborhood of a partition (N1 , N2 ) is the k-exchange
neighborhood . This is the set of all partitions obtained by selecting a fixed
number k of pairs of nodes (i, j) with i ∈ N1 and j ∈ N2 , and interchang-
ing them, that is, moving i into N2 and j into N1 . The corresponding local
search algorithm moves from a given solution to its minimum cost neighbor,
and terminates when no neighbor with smaller cost can be obtained. Unfor-
tunately, the amount of work needed to generate   a k-exchange neighborhood
increases exponentially with k [there are m k
different ways to select k objects
out of m]. One may thus consider a variable depth neighborhood that involves
multiple successive k-exchanges with small k. As an example, for k = 1 we
obtain the following algorithm:
Given the starting partition (N1 , N2 ), consider all pairs (i, j) with i ∈
N1 and j ∈ N2 , and let c(i, j) be the cost change that results from moving i
into N2 and j into N1 . If (i, j) is the pair that minimizes c(i, j), move i into
N1 and j into N2 , and let c1 = c(i, j). Repeat this process a fixed number M
of times, obtaining a sequence c2 , c3 , . . . , cM of minimal cost changes resulting
from the sequence of exchanges. Then find


m

m = arg min cl ,
m=1,...,M
l=1

and accept as the next partition the one involving the first m exchanges.
This type of algorithm avoids the exponential running time of k-exchange
neighborhoods, while still considering neighbors differing by as many as M
node pairs.

While the definition of neighborhood is often problem-dependent,

some general classes of procedures for generating neighborhoods have been
developed. One such class is genetic algorithms, to be discussed shortly.
In some cases, neighborhoods are dynamically changing, and they may de-
pend not only on the current solution, but also on several past solutions.
The method of tabu search, to be discussed shortly, falls in this category.
The criterion for selecting a solution from within a neighborhood is
usually the cost of the solution, but sometimes a more complex criterion
based on various problem characteristics and/or constraint violation con-
siderations is adopted. An important possibility, which is the basis for the
simulated annealing method , to be discussed shortly, is to use a random
mechanism for selecting the successor solution within a neighborhood.
Finally, regarding the termination criterion, many local search meth-
ods are cost improving, and stop when an improved solution cannot be
found within the current neighborhood. This means that these methods
stop at a local minimum, that is, a solution that is no worse than all other
solutions within its neighborhood. Unfortunately, for many problems, a
local minimum may be far from optimal, particularly if the neighborhood
used is relatively small. Thus, for a cost improving method, there is a basic
tradeoff between using a large neighborhood to diminish the difficulty with
514 Network Problems with Integer Constraints Chap. 10

local minima, and paying the cost of increased computation per iteration.
Note that there is an important advantage to a cost improving method: it
can never repeat the same solution, so that in view of the finiteness of the
feasible set F , it will always terminate with a local minimum.
An alternative type of neighbor selection and termination criterion,
used by simulated annealing and tabu search, is to allow successor solutions
to have worse cost than their predecessors, but to also provide mechanisms
that ensure the future generation of improved solutions with substantial
likelihood. The advantage of accepting solutions of worse cost is that stop-
ping at a local minimum becomes less of a difficulty. For example, the
method of simulated annealing, cannot be trapped at a local minimum,
as we will see shortly. Unfortunately, methods that do not enforce cost
improvement run the danger of cycling through repetition of the same so-
lution. It is therefore essential in these methods to provide a mechanism
by virtue of which cycling is either precluded, or becomes highly unlikely.
As a final remark, we note an important advantage of local search
methods. While they offer no solid guarantee of finding an optimal or
near-optimal solution, they offer the promise of substantial improvement
over any heuristic that can be used to generate the starting solution. Unfor-
tunately, however, one can seldom be sure that this promise will be fulfilled
in a given practical problem.

10.4.1 Genetic Algorithms

These are local search methods where the neighborhood generation mech-
anism is inspired by real-life processes of genetics and evolution. In par-
ticular, the current solution is modified by “splicing” and “mutation” to
obtain neighboring solutions. A typical example is provided by problems of
scheduling, such as the traveling salesman problem. The neighborhood of
a schedule T may be a collection of other schedules obtained by modifying
some contiguous portion of T in some way, while keeping the remainder of
the schedule T intact. Alternatively, the neighborhood of a schedule may
be obtained by interchanging the position of a few tasks, as in the k-OPT
traveling salesman heuristic.
In a variation of this approach, a pool of solutions may be maintained.
Some of these solutions may be modified, while some pairs of these solutions
may be combined to form new solutions. These solutions, are added to the
pool if some criterion, typically based on cost improvement, is met, and
some of the solutions of the existing pool may be dropped. In this way, it
is argued, the pool is “evolving” in a Darwinian way through a “survival
of the fittest” process.
A specific example implementation of this approach operates in phases.
At the beginning of a phase, we have a population X consisting of n feasible
solutions x1 , . . . , xn . The phase proceeds as follows:
Sec. 10.4 Local Search Methods 515

Typical Phase of a Genetic Algorithm

Step 1: (Local Search) Starting from each solution xi of the current

population X, apply a local search algorithm up to obtaining a local
minimum xi . Let X = {x1 , . . . , xn }.
Step 2: (Mutation) Select at random a subset of elements of X, and
modify each element according to some (problem dependent) mecha-
nism, to obtain another feasible solution.
Step 3: (Recombination) Select at random a subset of pairs of
elements of X, and produce from each pair a feasible solution according
to some (problem dependent) mechanism.
Step 4: (Survivor Selection) Let X̃ be the set of feasible solutions
obtained from the mutation and recombination Steps 3 and 4. Out of
the population X ∪ X̃, select a subset of n elements according to some
criterion. Use this subset to start the next phase.

Mutation allows speculative variations of the local minima at hand,

while recombination (also called crossover) aims to combine attributes of
a pair of local minima. The processes of mutation and recombination are
usually performed with the aid of some data structure that is used to rep-
resent a solution, such as for example a string of bits. There is a very large
number of variants of genetic algorithm approaches. Typically, these ap-
proaches are problem-dependent and require a lot of trial-and-error. How-
ever, genetic algorithms are quite easy to implement, and have achieved
considerable popularity. We refer to the literature cited at the end of the
chapter for more details.

10.4.2 Tabu Search

Tabu search aims to avoid getting trapped at a poor local minimum, by

accepting on occasion a worse or even infeasible solution from within the
current neighborhood. Since cost improvement is not enforced, tabu search
runs the danger of cycling, i.e., repeating the same sequence of solutions
indefinitely. To alleviate this problem, tabu search keeps track of recently
obtained solutions in a “forbidden” (tabu) list. Solutions in the tabu list
cannot be regenerated, thereby avoiding cycling, at least in the short run.
In a more sophisticated variation of this strategy, the tabu list contains
attributes of recently obtained solutions rather than the solutions them-
selves. Solutions with attributes in the tabu list are forbidden from being
generated (except under particularly favorable circumstances, under which
the tabu list is overridden).
Tabu search is also based on an elaborate web of implementation
heuristics that have been developed through experience with a large num-
516 Network Problems with Integer Constraints Chap. 10

ber of practical problems. These heuristics regulate the size of the current
neighborhood, the criterion of selecting a new solution from the current
neighborhood, the criterion for termination, etc. These heuristics may also
involve selective memory storage of previously generated solutions or their
attributes, penalization of the constraints with (possibly time-varying)
penalty parameters, and multiple tabu lists. We refer to the literature
cited at the end of the chapter for further details.

10.4.3 Simulated Annealing

Simulated annealing is similar to tabu search in that it occasionally allows

solutions of inferior cost to be generated. It differs from tabu search in the
manner in which it avoids cycling. Instead of checking deterministically
the preceding solutions for cycling, it simply randomizes its selection of the
next solution. In doing so, it not only avoids cycling, but also provides
some theoretical guarantee of escaping from local minima and eventually
finding a global minimum.
Being able to find a global minimum is not really exciting in itself.
For example, under fairly general conditions, one can do so by using unso-
phisticated random search methods, such as for example a method where
feasible solutions are sampled at random. However, simulated annealing
randomizes the choice of the successor solution from within the current
neighborhood in a way that gives preference to solutions of smaller cost,
and in doing so, it aims to find a global minimum faster than simple-minded
random search methods.
In particular, given a solution x, we select by random sampling a can-
didate solution x from the neighborhood N (x). The sampling probabilities
are positive for all members of N (x), but are otherwise unspecified. The
solution x is accepted if it is cost improving, that is

f (x) < f (x).

Otherwise, x is accepted with probability

 
e− f (x)−f (x) /T ,

where T is some positive constant, and is rejected with the complementary

probability.
The constant T regulates the likelihood of accepting solutions of worse
cost. It is called the temperature of the process (the name is inspired by a
certain physical analogy that will not be discussed here). The likelihood of
accepting a solution x of worse cost than x decreases as its cost increases.
Furthermore, when T is large (or small), the probability of accepting a
worse solution is close to 1 (or close to 0, respectively). In practice, it is
typical to start with a large T , allowing a better chance of escaping from
 Sec. 10.5 Rollout Algorithms 517

local minima, and then to reduce T gradually to enhance the selectivity of

the method towards improved solutions.
Contrary to genetic algorithms and tabu search, which offer no gen-
eral theoretical guarantees of good performance, simulated annealing is
supported by solid theory. In particular, under fairly general conditions,
it can be shown that a global minimum will be eventually visited (with
probability 1), and that with gradual reduction of the temperature T , the
search process will be confined with high likelihood to solutions that are
globally optimal.
For an illustrative analysis, assume that T is kept constant and let
pxy be the probability that when the current solution is x, the next solu-
tion sampled is y. Consider the special case where pxy = pyx for all feasible
solutions x and y, and assume that the Markov chain defined by the prob-
abilities pxy is irreducible, in the sense that there is positive probability to
go from any x to any y, with one or more samples. Then it can be shown
(see Exercise 10.20) that the steady-state probability of a solution x is

e−f (x)/T
−f (x)/T
.
x∈F e

Essentially, this says that for very small T and far into the future, the
current solution is almost always optimal.
When the condition pxy = pyx does not hold, one cannot obtain a
closed-form expression for the steady-state probabilities of the various so-
lutions. However, as long as the underlying Markov chain is irreducible,
the behavior is qualitatively similar: the steady-state probability of nonop-
timal solutions diminishes to 0 as T approaches 0. There is also related
analysis for the case where the temperature parameter T is time-varying
and converges to 0; see the references cited at the end of the chapter.
The results outlined above should be viewed with a grain of salt. In
practice, speed of convergence is as important as eventual convergence to
the optimum, and solving a given problem by simulated annealing can be
very slow. A nice aspect of the method is that it depends very little on
the structure of the problem being solved, and this enhances its value for
relatively unstructured problems that are not well-understood. For other
problems, where there exists a lot of accumulated insight and experience,
simulated annealing is usually inferior to other local search approaches.

10.5 ROLLOUT ALGORITHMS

The branch-and-bound algorithm is guaranteed to find an optimal flow

vector, but it may require the solution of a very large number of sub-
problems. Basically, the algorithm amounts to an exhaustive search of the
518 Network Problems with Integer Constraints Chap. 10

entire branch-and-bound tree. An alternative is to consider faster methods

that are based on intelligent but nonexhaustive search of the tree. In this
section, we develop one such method, the rollout algorithm, which, in its
simplest version, sequentially constructs a suboptimal flow vector by fixing
the arc flows, a few arcs at a time. The rollout algorithm can be combined
with most heuristics, including the local search methods of the preceding
section, and is capable of magnifying their effectiveness.
Let us consider the minimization of a function f of a flow vector x
over a feasible set F , and let us assume that F is finite (presumably because
of some integer constraints on the arc flows). Define a partial solution to
be a collection of arc flows {xij | (i, j) ∈ S}, corresponding to some proper
subset of arcs S ⊂ A. Such a collection is distinguished from a flow vector
(S = A), which is also referred to as a complete solution.
The rollout algorithm generates a sequence of partial solutions, cul-
minating with a complete solution. For this purpose, it employs some
problem-dependent heuristic algorithm, called the base heuristic. This al-
gorithm, given a partial solution


P = xij | (i, j) ∈ S ,

produces a complementary solution


P = xij | (i, j) ∈
/S ,

and a corresponding (complete) flow vector


x = xij | (i, j) ∈ A = P ∪ P .

The cost of this flow vector is denoted by


f (x) if x ∈ F ,
H(P ) =
∞ otherwise,

and is called the heuristic cost of the partial solution P . If P is a com-

plete solution, which is feasible, i.e., a flow vector x ∈ F , by convention the
heuristic cost of P is the true cost f (x). There are no restrictions on the na-
ture of the base heuristic; a typical example is an integer rounding heuristic
applied to the solution of some related linear or convex network problem,
which may be obtained by relaxing/neglecting the integer constraints.
The rollout algorithm starts with some partial solution, or with the
empty set of arcs, S = Ø. It enlarges a partial solution iteratively, with a
few arc flows at a time. The algorithm terminates when a complete solution
is obtained. At the start of the typical iteration, we have a current partial
solution

P = xij | (i, j) ∈ S ,
Sec. 10.5 Rollout Algorithms 519

and at the end of the iteration, we augment this solution with some more
arc flows. The steps of the iteration are as follows:

Iteration of the Rollout Algorithm

Step 1: Select a subset T of arcs that are not in S according to some

criterion. (The arc selection method is usually based on some heuristic
preliminary optimization, and is problem-dependent.)
Step 2: Consider

the collection
 FT of all possible values of the arc
flows y = yij | (i, j) ∈ T , and apply the base heuristic to compute
the heuristic cost H(Py+ ) of the augmented partial solution




Py+ = xij | (i, j) ∈ S , yij | (i, j) ∈ T

for each y ∈ FT .


Step 3: Choose from the set FT the arc flows y = y ij | (i, j) ∈ T
that minimize the heuristic cost H(Py+ ); that is, find

y = arg min H(Py+ ). (10.35)

y∈FT


Step 4: Augment

the currentpartial solution {xij | (i, j) ∈ S with
the arc flows y ij | (i, j) ∈ T thus obtained, and proceed with the
next iteration.

As an example of this algorithm, let us consider the traveling salesman

problem, and let us use as base heuristic the nearest neighbor method,
whereby we start from some simple path and at each iteration, we add a
node that does not close a cycle and minimizes the cost of the enlarged path.
The rollout algorithm operates as follows: After k iterations, we have a path
{i1 , . . . , ik } consisting of distinct nodes. At the next iteration, we run the
nearest neighbor heuristic starting from each of the paths {i1 , . . . , ik , i}
with i = i1 , . . . , ik , and obtain a corresponding tour. We then select as
next node ik+1 of the path the node i that corresponds to the best tour
thus obtained. Here, the set of arcs used to augment the current partial
solution in the rollout algorithm is


T = (ik , i) | i = i1 , . . . , ik ,

and at the kth iteration the flows of all of these arcs are set to 0, except
for arc (ik , ik+1 ) whose flow is set to 1.
Note that a rollout algorithm requires considerably more computation
than the base heuristic. For example, in the case where the subset T in Step
520 Network Problems with Integer Constraints Chap. 10

1 consists of a single arc, the rollout algorithm requires O(mn) applications

of the base heuristic, where
m is the number of arcs, and
n is a bound on the number of possible values of each arc flow.
Nonetheless the computational requirements of the rollout algorithm may
be quite manageable. In particular, if the arc flows are restricted to be
0 or 1, and the base heuristic has polynomial running time, so does the
corresponding rollout algorithm.
An important question is whether, given an initial partial solution,
the rollout algorithm performs at least as well as its base heuristic when
started from that solution. This can be guaranteed if the base heuristic is
sequentially consistent. By this we mean that the heuristic has the following
property:
Suppose that starting from a partial solution


P = xij | (i, j) ∈ S ,

the heuristic produces the complementary solution


P = xij | (i, j) ∈
/S .

Then starting from the partial solution


P + = xij | (i, j) ∈ S ∪ T ,

the heuristic produces a complementary solution

+

P = xij | (i, j) ∈
/ S∪T ,

which coincides with P on the arcs (i, j) ∈

/ S ∪ T.
As an example, it can be seen that the nearest neighbor heuristic for
the traveling salesman problem, discussed earlier, is sequentially consistent.
This is a manifestation of a more general property: many common base
heuristics of the greedy type are by nature sequentially consistent (see
Exercise 10.21). It may be verified, based on Eq. (10.35), that a sequentially
consistent rollout algorithm keeps generating the same solution P ∪ P , up
to the point where by examining the alternatives in Eq. (10.35) and by
calculating their heuristic costs, it discovers a better solution. As a result,
sequential consistency guarantees that the costs of the successive solutions
P ∪ P produced by the rollout algorithm are monotonically nonincreasing;
that is, we have
H(P + ) ≤ H(P )
at every iteration. Thus, the cost f (xt ) of the solution xt produced upon
termination of the rollout algorithm is at least as small as the cost f (x0 )
Sec. 10.5 Rollout Algorithms 521

of the initial solution x0 produced by the base heuristic. For further elab-
oration of the sequential consistency property, we refer to the paper by
Bertsekas, Tsitsiklis, and Wu [1997], which also discusses some underlying
connections with the policy iteration method of dynamic programming.
A condition that is more general than sequential consistency is that
the algorithm be sequentially improving, in the sense that at each iteration
there holds
H(P + ) ≤ H(P ).
This property also guarantees that the cost of the solutions produced by the
rollout algorithm is monotonically nonincreasing. The paper by Bertsekas,
Tsitsiklis, and Wu [1997] discusses situations where this property holds,
and shows that with fairly simple modification, a rollout algorithm can be
made sequentially improving (see also Exercise 10.22).
There are a number of variations of the basic rollout algorithm de-
scribed above. Here are some examples:
(1) We may adapt the rollout framework to use multiple heuristic al-
gorithms. In particular, let us assume that we have K algorithms
H1 , . . . , HK . The kth of these algorithms, given an augmented par-
tial solution Py+ , produces a heuristic cost Hk (Py+ ). We may then use
in the flow selection via Eq. (10.35) a heuristic cost of the form

H(Py+ ) = min Hk (Py+ ),

k=1,...,K

or of the form

K
H(Py+ ) = rk Hk (Py+ ),
k=1

where rk are some fixed scalar weights obtained by trial and error.
(2) We may incorporate multistep lookahead or selective depth lookahead
into the rollout framework.
Here we consider
 augmenting the current
partial solution P = xij | (i, j) ∈ S with all possible values for
the flows of a finite sequence of arcs that are not in S. We run the
base heuristic from each of the corresponding augmented partial so-
lutions, we select the sequence of arc flows with minimum heuristic
cost, and then augment the current partial solution P with the first
arc flow in this sequence. As an illustration, let us recall the traveling
salesman problem with the nearest neighbor method used as the base
heuristic. An example rollout algorithm with two-step lookahead op-
erates as follows: We begin each iteration with a path {i1 , . . . , ik }.
We run the nearest neighbor heuristic starting from each of the paths
{i1 , . . . , ik , i} with i = i1 , . . . , ik , and obtain a corresponding tour.
We then form the subset I consisting of the m nodes i = i1 , . . . , ik
that correspond to the m best tours thus obtained. We run the near-
est neighbor heuristic starting from each of the paths {i1 , . . . , ik , i, j}
522 Network Problems with Integer Constraints Chap. 10

with i ∈ I and j = i1 , . . . , ik , i, and obtain a corresponding tour. We

then select as the next node ik+1 of the path the node i ∈ I that
corresponds to a minimum cost tour.
(3) We may use alternative methods for computing a cost H(Py+ ) of a
candidate augmented partial solution Py+ for use in the flow selection
via Eq. (10.35). For example, instead of generating this cost via the
base heuristic, we may calculate it as the optimal or approximately
optimal cost of a suitable optimization problem. In particular, it is
possible to use a cost derived from Lagrangian relaxation, whereby
at a given partial solution, an appropriate dual problem is solved,
and its optimal cost is used in place of the heuristic cost H in Eq.
(10.35). Alternatively, a complementary solution may be constructed
based on minimization of the corresponding Lagrangian function. As
another example, one may use as cost of a partial solution, some
heuristic measure of quality of the partial solution; this idea forms
the basis for computer chess, where various positions are evaluated
using a heuristic “position evaluation function.”
Let us provide a few examples of rollout algorithms. The first example
is very simple, but illustrates well the notions of sequential consistency and
sequential improvement.

Example 10.13. (One-Dimensional Walk)

Consider a person who walks on a straight line and at each time period takes
either a unit step to the left or a unit step to the right. There is a cost
function assigning cost f (i) to each integer i. Given an integer starting point
on the line, the person wants to minimize the cost of the point where he will
end up after a given and fixed number N of steps.
We can formulate this problem as a problem of selecting a path in a
graph (see Fig. 10.15). In particular, without loss of generality, let us assume
that the starting point is the origin, so that the person’s position after N
steps will be some integer in the interval [−N, N ]. The nodes of the graph
are identified with pairs (k, m), where k is the number of steps taken so far
(k = 1, . . . , N ) and m is the person’s position (m ∈ [−k, k]). A node (k, m)
with k < N has two outgoing arcs with end nodes (k+1, m−1) (corresponding
to a left step) and (k+1, m+1) (corresponding to a right step). Let us consider
paths whose starting node is (0, 0) and the destination node is of the form
(N, m), where m is of the form N − 2l and l ∈ [0, N ] is the number of left
steps taken. The problem then is to find the path of this type such that f (m)
is minimized.
Let the base heuristic be the algorithm, which, starting at a node (k, m),
takes N − k successive steps to the right and terminates at the node (N, m +
N − k). It can be seen that this algorithm is sequentially consistent [the base
heuristic generates the path (k, m), (k + 1, m + 1), . . . , (N, m + N − k) starting
from (k, m), and also the path (k + 1, m + 1), . . . , (N, m + N − k) starting
from (k + 1, m + 1), so the criterion for sequential consistency is fulfilled].
Sec. 10.5 Rollout Algorithms 523

The rollout algorithm, at node (k, m) compares the cost of the des-
tination node (N, m + N − k) (corresponding to taking a step to the right
and then following the base heuristic) and the cost of the destination node
(N, m + N − k − 2) (corresponding to taking a step to the left and then fol-
lowing the base heuristic). Let us say that an integer i ∈ [−N + 2, N − 2] is
a local minimum if f (i − 2) ≥ f (i) and f (i) ≤ f (i + 2). Let us also say that
N (or −N ) is a local minimum if f (N − 2) ≤ f (N ) [or f (−N ) ≤ f (−N + 2),
respectively]. Then it can be seen that starting from the origin (0, 0), the
rollout algorithm obtains the local minimum that is closest to N , (see Fig.
10.15). This is no worse (and typically better) than the integer N obtained
by the base heuristic. This example illustrates how the rollout algorithm may
exhibit “intelligence” that is totally lacking from the base heuristic.

(0,0)

Time

(N,-N ) (N,0) (N,N )

i
f (i ) Position

-N 0 i N-2 N i

Figure 10.15: Illustration of the path generated by the rollout algorithm in

Example 10.13. It keeps moving to the left up to the time where the base
heuristic generates two destinations (N, i) and (N, i − 2) with f (i) ≤ f (i − 2).
Then it continues to move to the right ending at the destination (N, i), which
corresponds to the local minimum closest to N .

Consider next the case where the base heuristic is the algorithm that,
starting at a node (k, m), compares the cost f (m + N − k) (corresponding to
taking all of the remaining N −k steps to the right) and the cost f (m−N +k)
(corresponding to taking all of the remaining N − k steps to the left), and
accordingly moves to node

(N, m + N − k) if f (m + N − k) ≤ f (m − N + k),
524 Network Problems with Integer Constraints Chap. 10

or to node

(N, m − N + k) if f (m − N + k) < f (m + N − k).

It can be seen that this base heuristic is not sequentially consistent, but is
instead sequentially improving. It can then be verified that starting from the
origin (0, 0), the rollout algorithm obtains the global minimum of f in the
interval [−N, N ], while the base heuristic obtains the better of the two points
−N and N .

Example 10.14. Constrained Traveling Salesman Problem

Consider the traveling salesman problem of Example 10.1, where we want to

minimize the cost 
aij ,
(i,j)∈T

of a tour T , while satisfying the side constraints


ckij ≤ dk , ∀ k = 1, . . . , K.
(i,j)∈T

A rollout algorithm starts with the trivial path P = (s), where s is some
initial node, progressively constructs a sequence of paths P = (s, i1 , . . . , im ),
m = 1, . . . , N − 1, consisting of distinct nodes, and then completes a tour by
adding the arc (iN −1 , s). The rollout procedure is as follows.
We introduce nonnegative penalty coefficients µk for the side constraints,
and we form modified arc traversal costs âij , given by


K

âij = aij + µk ckij .

k=1

The method of obtaining µk is immaterial for our purposes in this example,

but we note that one possibility is to use the Lagrangian relaxation method
of Section 10.3. We assume that we have a heuristic algorithm that can
complete the current path P = (s, i1 , . . . , im ) with a path (im+1 , . . . , iN −1 , s),
thereby obtaining a tour T ∗ (P ) that has approximately minimum modified
cost. Some of the heuristics mentioned in Example 10.1, including the k-OPT
heuristic, can be used for this purpose. Furthermore, we assume that by using
another heuristic, we can complete the current path P to a tour T̂ (P ) that
satisfies all the side constraints.
Given the current path P = (s, i1 , . . . , im ), the rollout algorithm, con-
siders the set Am of all arcs (im , j) ∈ A such that j does not belong to P .
For each of the nodes j such that (im , j) ∈ Am , it considers the expanded
path Pe = (s, i1 , . . . , im , j) and obtains the tours T ∗ (Pe ) and T̂ (Pe ), using the
heuristics mentioned earlier. The rollout algorithm then adds to the current
partial path P the node j for which the tour T ∗ (Pe ) satisfies the side con-
straints and has minimum cost (with respect to the arc costs aij ); if no path
 Sec. 10.6 Notes, Sources, and Exercises 525

T ∗ (Pe ) satisfies the side constraints, the algorithm adds to the current path
the node j for which the tour T̂ (Pe ) has minimum cost.
One of the drawbacks of the scheme just described is that it requires the
approximate solution of a large number of traveling salesman problems. A
faster variant is obtained if the arc set Am above is restricted to be a suitably
chosen subset of the eligible arcs (im , j), such for example those whose length
does not exceed a certain threshold.

Finally, it is interesting to compare rollout algorithms with the local

search methods of the preceding section. Both types of algorithms generate
a sequence of solutions, but in the case of a rollout algorithm, the generated
solutions are partial (except at termination), while in a local search method,
the generated solutions are complete. In both types of algorithms, the next
solution is generated from within a neighborhood of the current solution,
but the selection criterion in rollout algorithms is the estimated cost of the
solution as obtained by the base heuristic, while in local search methods,
it is typically the true cost of the solution. Finally, in rollout algorithms,
there is no concern about local minima and cycling, but there is also no
provision for improving a complete solution after it is obtained.
There are interesting possibilities for combining a rollout algorithm
with a local search method. In particular, one may use a local search
method as part of a base heuristic in a rollout algorithm; here, the local
search method could be fairly unsophisticated, since one may hope that
the rollout process will provide an effective mechanism for solution im-
provement. Alternatively, one may first use a rollout algorithm to obtain
a complete solution, and then use a local search method in an effort to
improve this solution.

10.6 NOTES, SOURCES, AND EXERCISES

There is a great variety of integer constrained network flow problems, and

the associated methodological and applications literature is vast. For text-
book treatments at various levels of sophistication, which also cover broader
aspects of integer programming, see Lawler [1976], Zoutendijk [1976], Pa-
padimitriou and Steiglitz [1982], Minoux [1986a], Schrijver [1986], Nem-
hauser and Wolsey [1988], Bogart [1990], Pulleyblank, Cook, Cunning-
ham, and Schrijver [1993], Cameron [1994], and Cook, Cunningham, Pul-
leyblank, and Schrijver [1998]. Volumes 7 and 8 of the Handbooks in
Operations Research and Management Science, edited by Ball, Magnanti,
Monma, and Nemhauser [1995a], [1995b], are devoted to network theory
and applications, and include several excellent survey papers with large bib-
liographies. O’hEigeartaigh, Lenstra, and Rinnoy Kan [1985] provide an
extensive bibliography on combinatorial optimization. Von Randow [1982],
526 Network Problems with Integer Constraints Chap. 10

[1985] gives an extensive bibliography on integer programming and related

subjects.
The traveling salesman problem has been associated with many of the
important investigations in discrete optimization. It was first considered
in a modern setting by Dantzig, Fulkerson, and Johnson [1954], whose pa-
per stimulated much interest and research. The edited volume by Lawler,
Lenstra, Rinnoy Kan, and Shmoys [1985] focuses on the traveling sales-
man problem and its variations, and the papers by Junger, Reinelt, and
Rinaldi [1995], and by Johnson and McGeoch [1997] provide extensive sur-
veys of the subject. There is a large literature on the use of polyhedral
approximations to the feasible set of integer programming problems and
the traveling salesman problem in particular; see, for example, the papers
by Cornuejols, Fonlupt, and Naddef [1985], Grötschel and Padberg [1985],
Padberg and Grötschel [1985], Pulleyblank [1983], and the books by Nem-
hauser and Wolsey [1988], and Schrijver [1986]. The papers by Burkard
[1990], Gilmore, Lawler, and Shmoys [1985], and Tsitsiklis [1992] discuss
some special cases of the traveling salesman problem and some extensions.
The monograph by Martello and Toth [1990] is devoted to generalized
assignment problems, including ones with integer constraints. The book
by Kershenbaum [1993] provides a lot of material on tree construction and
network design algorithms for data communications; see also Monma and
Sheng [1986], Minoux [1989], Bertsekas and Gallager [1992], and Grötschel,
Monma, and Stoer [1995]. Exact and heuristic methods for the Steiner tree
problem are surveyed by Winter [1987] and Voß [1992].
Matching problems are discussed in detail in the monograph by Lovasz
and Plummer [1985], the survey by Gerards [1995], and Chapter 10 of
the book by Murty [1992]. For vehicle and arc routing problems, see the
surveys by Assad and Golden [1995], Desrosiers, Dumas, Solomon, and
Soumis [1995], Eiselt, Gendreau, and Laporte [1995a], [1995b], Federgruen
and Simchi-Levi [1995], Fisher [1995], and Powell, Jaillet, and Odoni [1995].
An important application of multidimensional assignment problems
arises in the context of multi-target tracking and data association; see
Blackman [1986], Bar-Shalom and Fortman [1988], Pattipati, Deb, Bar-
Shalom, and Washburn [1992], Poore [1994], Poore and Robertson [1997].
The material on the error bounds for the enforced separation heuristic in
three-dimensional assignment problems (Exercise 10.31) is apparently new.
Integer multicommodity flow problems are discussed by Barnhart,
Hane, and Vance [1997]. Nonlinear, nonconvex network optimization is
discussed by Lamar [1993], Bell and Lamar [1993], as well as in general texts
on global optimization; see Pardalos and Rosen [1987], Floudas [1995], and
Horst, Pardalos, and Thoai [1995]. For a textbook treatment of scheduling
(cf. Exercises 10.23-10.27), see Pinedo [1995].
Branch-and-bound has its origins in the traveling salesman paper by
Dantzig, Fulkerson, and Johnson [1954]. Their paper was followed by Croes
[1958], Eastman [1958], and Land and Doig [1960], who considered versions
Sec. 10.6 Notes, Sources, and Exercises 527

of the branch-and-bound method in the context of various integer program-

ming problems. The term “branch-and-bound” was first used by Little,
Murty, Sweeney, and Karel [1963], in the context of the traveling salesman
problem. Balas and Toth [1985], and Nemhauser and Wolsey [1988] provide
extensive surveys of branch-and-bound.
Lagrangian relaxation was suggested in the context of discrete op-
timization by Held and Karp [1970], [1971]. Subgradient methods were
introduced by Shor in the Soviet Union during the middle 60s. The conver-
gence properties of subgradient methods and their variations are discussed
in a number of sources, including Auslender [1976], Goffin [1977], Shapiro
[1979], Shor [1985], Poljak [1987], Hiriart-Urruty and Lemarechal [1993],
Brannlund [1993], Bertsekas [1995b], and Goffin and Kiwiel [1996].
Cutting plane methods were proposed by Cheney and Goldstein [1959],
and by Kelley [1960]; see also the book by Goldstein [1967]. Central cut-
ting plane methods were introduced by Elzinga and Moore [1975]. More
recent proposals, some of which relate to interior point methods, are dis-
cussed in Goffin and Vial [1990], Goffin, Haurie, and Vial [1992], Ye [1992],
Kortanek and No [1993], Goffin, Luo, and Ye [1993], [1996], Atkinson and
Vaidya [1995], Nesterov [1995], Luo [1996], and Kiwiel [1997b].
Three historically important references on decomposition methods
are Dantzig and Wolfe [1960], Benders [1962], and Everett [1963]. An
early text on large-scale optimization and decomposition is Lasdon [1970];
see also Geoffrion [1970], [1974]. Subgradient methods have been applied
to the solution of multicommodity flow problems using a decomposition
framework by Kennington and Shalaby [1977]. The book by Censor and
Zenios [1997] discusses several applications of decomposition in a variety of
algorithmic contexts.
The literature of local search methods is extensive. The edited volume
by Aarts and Lenstra [1997] contains several surveys of broad classes of
methods. Osman and Laporte [1996] provide an extensive bibliography.
The book by Goldberg [1989] focuses on genetic algorithms. Tabu
search was initiated with the works of Glover [1986] and Hansen [1986].
The book by Glover and Laguna [1997], and the surveys by Glover [1989],
[1990], Glover, Taillard, and de Verra [1993] provide detailed expositions
and give many references.
Simulated annealing was proposed by Kirkpatrick, Gelatt, and Vecchi
[1983] based on earlier suggestions by Metropolis, Rosenbluth, Rosenbluth,
Teller, and Teller [1953]; see also Cerny [1985]. The main theoretical con-
vergence properties of the method were established by Hajek [1988] and
Tsitsiklis [1989]; see also the papers by Connors and Kumar [1989], Gelfand
and Mitter [1989], and Bertsimas and Tsitsiklis [1993], and the book by Ko-
rst, Aarts, and Korst [1989]. A framework for integration of local search
methods is presented by Fox [1993], [1995].
Rollout algorithms for discrete optimization were proposed in the
book by Bertsekas and Tsitsiklis [1996] in the context of the neuro-dynamic
528 Network Problems with Integer Constraints Chap. 10

programming methodology, and in the paper by Bertsekas, Tsitsiklis, and

Wu [1997]. An application to scheduling using the framework of the quiz
problem (cf. Exercises 10.28 and 10.29) is described by Bertsekas and
Castañon [1998]. The idea of sequential selection of candidates for par-
ticipation in a solution is implicit in several combinatorial optimization
contexts. For example this idea is embodied in the sequential fan candi-
date list strategy as applied in tabu search (see Glover, Taillard, and de
Werra [1993]). A similar idea is also used in the sequential automatic test
procedures of Pattipati (see e.g., Pattipati and Alexandridis [1990]).

EXERCISES

10.1

Consider the symmetric traveling salesman problem with the graph shown in Fig.
10.16.
(a) Find a suboptimal solution using the nearest neighbor heuristic starting
from node 1.
(b) Find a suboptimal solution by first solving an assignment problem, and by
then merging subtours.
(c) Try to improve the solutions found in (a) and (b) by using the 2-OPT
heuristic.

5
2 5
9 9
1 0 4
Figure 10.16: Data for a symmetric trav-
1 2 eling salesman problem (cf. Exercise 10.1).
8 8 The arc costs are shown next to the arcs.
2 3 Each arc is bidirectional.
5

Symmetric Traveling Salesman

Problem Data.
Costs Shown Next to the Arcs.
Each arc is bidirectional.
Sec. 10.6 Notes, Sources, and Exercises 529

10.2 (Minimum Cost Cycles)

Consider a strongly connected graph with a nonnegative cost for each arc. We
want to find a forward cycle of minimum cost that contains all nodes but is not
necessarily simple; that is, a node or an arc may be traversed multiple times.
(a) Convert this problem into a traveling salesman problem. Hint: Construct
a complete graph with cost of an arc (i, j) equal to the shortest distance
from i to j in the original graph.
(b) Apply your method of part (a) to the graph of Fig. 10.17.

2
3 2
Figure 10.17: Data for a minimum cost cycle
1 1 4 problem (cf. Exercise 10.2). The arc costs are
1 1
shown next to the arcs.
1 5
3

10.3

Consider the problem of checking whether a given graph contains a simple cycle
that passes through all the nodes. (The cycle need not be forward.) Formulate
this problem as a symmetric traveling salesman problem. Hint: Consider a
complete graph where the cost of an arc (i, j) is 1 if (i, j) or (j, i) is an arc
of the original graph, and is 2 otherwise.

10.4

Show that an asymmetric traveling salesman problem with nodes 1, . . . , N and arc
costs aij can be converted to a symmetric traveling salesman problem involving
a graph with nodes 1, . . . , N, N + 1, . . . , 2N, and the arc costs

aij if i, j = 1, . . . , N, i = j,
ai(N +j) =
−M if i = j,

where M is a sufficiently large number. Hint: All arcs with cost −M must be
included in an optimal tour of the symmetric version.

10.5

Consider the problem of finding a shortest (forward) path from an origin node s
to a destination node t of a graph with given arc lengths, subject to the additional
constraint that the path passes through every node exactly once.
530 Network Problems with Integer Constraints Chap. 10

(a) Show that the problem can be converted to a traveling salesman problem
by adding an artificial arc (t, s) of length −M , where M is a sufficiently
large number.
(b) (Longest Path Problem) Consider the problem of finding a simple forward
path from s to t that has a maximum number of arcs. Show that the
problem can be converted to a traveling salesman problem.

10.6

Consider the problem of finding a shortest (forward) path in a graph with given
arc lengths, subject to the constraint that the path passes through every node
exactly once (the choice of start and end nodes of the path is subject to opti-
mization). Formulate the problem as a traveling salesman problem.

10.7 (Traveling Salesman Problem/Triangle Inequality)

Consider a symmetric traveling salesman problem where the arc costs are non-
negative and satisfy the following triangle inequality:

aij ≤ aik + akj , for all nodes i, j, k.

This problem has some special algorithmic properties.

(a) Consider a procedure, which given a cycle {i0 , i1 , . . . , iK , i0 } that contains
all the nodes (but passes through some of them multiple times), obtains a
tour by deleting nodes after their first appearance in the cycle; e.g., in a
5-node problem, starting from the cycle {1, 3, 5, 2, 3, 4, 2, 1}, the procedure
produces the tour {1, 3, 5, 2, 4, 1}. Use the triangle inequality to show that
the tour thus obtained has no greater cost than the original cycle.
(b) Starting with a spanning tree of the graph, use the procedure of part (a)
to construct a tour with cost equal to at most two times the total cost of
the spanning tree. Hint: The cycle should cross each arc of the spanning
tree exactly once in each direction. “Double” each arc of the spanning tree.
Use the fact that if a graph is connected and each of its nodes has even
degree, there is a cycle that contains all the arcs of the graph exactly once
(cf. Exercise 1.5).
(c) (Double tree heuristic) Start with a minimum cost spanning tree of the
graph, and use part (b) to construct a tour with cost equal to at most
twice the optimal tour cost.
(d) Verify that the problem of Fig. 10.18 satisfies the triangle inequality. Apply
the method of part (c) to this problem.

10.8 (Christofides’ Traveling Salesman Heuristic)

Consider a symmetric traveling salesman problem where the arc costs are non-
negative and satisfy the triangle inequality (cf. the preceding exercise). Let R
Sec. 10.6 Notes, Sources, and Exercises 531

5
2 5
3 6
1 3 4
Figure 10.18: Data for a symmetric travel-
1 2 ing salesman problem (cf. Exercises 10.7 and
4 4 10.8). The arc costs are shown next to the
2 3
5 arcs.

Symmetric Traveling Salesman

Problem Data.
Costs Shown Next to the Arcs.
Each arc is bidirectional.

be a minimum cost spanning tree of the graph (cf. Exercise 2.30), and let S be
the subset of the nodes that has an odd number of incident arcs in R. A perfect
matching of the nodes of S is a subset of arcs such that every node of S is an end
node of exactly one arc of the subset and each arc of the subset has end nodes
in S. Suppose that M is a perfect matching of the nodes of S that has minimum
sum of arc costs. Construct a tour that consists of the arcs of M and some of the
arcs of R, and show that its weight is no more than 3/2 times the optimal tour
cost. Solve the problem of Fig. 10.18 using this heuristic, and find the ratio of
the solution cost to the optimal tour cost. Hint: Note that the total cost of the
arcs of M is at most 1/2 the optimal tour cost. Also, use the fact that if a graph
is connected and each of its nodes has even degree, there is a cycle that contains
all the arcs of the graph exactly once (cf. Exercise 1.5).

10.9 (K-Traveling Salesmen Problem)

Consider the version of the traveling salesman problem where there are K sales-
men that start at city 1, return to city 1, and collectively must visit all other
cities exactly once. Transform the problem into an ordinary traveling salesman
problem. Hint: Split city 1 into K cities.

10.10 (Degree-Constrained Minimum Weight Spanning Trees)

Consider the minimum weight spanning tree problem, subject to the additional
constraint that the number of tree arcs that are incident to a single given node
s should be no greater than a given integer k. Consider adding a nonnegative
weight w to the weight of all incident arcs of node s, solving the corresponding
unconstrained spanning tree problem, and gradually increasing w until the degree
constraint is satisfied.
(a) State a polynomial algorithm for doing this and derive its running time.
(b) Use this algorithm to solve the problem of Fig. 10.19, where the degree of
node 1 is required to be no more than 2.
532 Network Problems with Integer Constraints Chap. 10

5
2 9
1
1 4
Figure 10.19: Data for a minimum weight
9 5
spanning tree problem (cf. Exercises 10.10 and
1 2 10.11). The arc weights are shown next to the
5 arcs.
2 3

Spanning Tree Problem Data.

Weights Shown Next to the Arcs.

10.11 (Steiner Tree Problem Heuristic)

We are given a connected graph G with a nonnegative weight aij for each arc
(i, j) ∈ A. We assume that if an arc (i, j) is present, the reverse arc (j, i) is also
present, and aij = aji . Consider the problem of finding a tree in G that spans a
given subset of nodes S and has minimum weight over all such trees.
(a) Let W ∗ be the weight of this tree. Consider the graph I(G), which has
node set S and is complete (has an arc connecting every pair of its nodes).
Let the weight for each arc (i, j) of I(G) be equal to the shortest distance in
the graph G from the node i ∈ S to the node j ∈ S. Let T be a minimum
weight spanning tree of I(G). Show that the weight of T is no greater
than 2W ∗ . Hint: Consider a minimum weight tour in I(G). Show that the
weight of this tour is no less than the weight of T and no more than 2W ∗ .
(b) Construct a heuristic based on part (a) and apply it to the problem of Fig.
10.19, where S = {1, 3, 5}.

10.12 (A General Heuristic for Spanning Tree Problems)

Consider a minimum weight spanning tree problem with an additional side con-
straint denoted by C (for example, a degree constraint on each node). A general
heuristic (given by Deo and Kumar [1997]) is to solve the problem neglecting
the constraint C, and then to add a scalar penalty to the cost of the arcs that
“contribute most” to violation of C. This is then repeated as many times as
desired.
(a) Construct a heuristic of this type for the capacitated spanning tree problem
(cf. Example 10.3).
(b) Adapt this heuristic to a capacitated Steiner tree problem.

10.13

Consider the Königsberg bridge problem (cf. Fig. 10.6).

Sec. 10.6 Notes, Sources, and Exercises 533

(a) Suppose that there existed a second bridge connecting the islands B and
C, and also another bridge connecting the land areas A and D. Construct
an Euler cycle that crosses each of the bridges exactly once.
(b) Suppose the bridge connecting the islands B and C has collapsed. Con-
struct an Euler path, i.e., a path (not necessarily a cycle) that passes
through each arc of the graph exactly once.
(c) Construct an optimal postman cycle assuming all arcs have cost 1.

10.14

Formulate the capacitated spanning tree problem given in Fig. 10.5 as an integer-
constrained network flow problem.

10.15 (Network Formulation of Nonbipartite Matching)

Consider the nonbipartite matching problem of Example 10.4. Replace each

node i with a pair of nodes i and i . For every arc (i, j) of the original problem,
introduce an arc (i, j  ) with value aij and an arc (j, i ) also with value aij . Show
that the problem can be formulated as the assignment-like problem involving the
conservation of flow inequalities

xij  ≤ 1, ∀ i,
j


xij  ≤ 1, ∀ j,
i

the integer constraints xij  ∈ {0, 1}, and the side constraints
 
xij + xji ≤ 1, ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

or  
xij + xji = 1, ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

in the case where a perfect matching is sought.

10.16 (Matching Solution of the Chinese Postman Problem)

Given a Chinese postman problem, delete all nodes of even degree together with
all their incident arcs. Find a perfect matching of minimum cost in the remaining
graph. Create an expanded version of the original problem’s graph by adding an
extra copy of each arc of the minimum cost matching. Show that an Euler cycle
of the expanded graph is an optimal solution to the Chinese postman problem.
534 Network Problems with Integer Constraints Chap. 10

10.17 (Solution of the Directed Chinese Postman Problem)

Consider expanding the graph of the directed Chinese postman problem by du-
plicating arcs so that the number of incoming arcs to each node is equal to the
number of its outgoing arcs. A forward Euler cycle of the expanded graph cor-
responds to a solution of the directed Chinese postman problem. Show that the
optimal expanded graph is obtained by minimizing

aij xij
(i,j)∈A

subject to the constraints

 
xij − xji = di , ∀ i ∈ N,
{j|(i,j)∈A} {j|(j,i)∈A}

0 ≤ xij , ∀ (i, j) ∈ A,

where di is the difference between the number of incoming arcs to i and the
number of outgoing arcs from i.

10.18 (Shortest Paths and Branch-and-Bound)

Consider a general integer-constrained problem of the form

minimize f (x1 , . . . , xn )
subject to x ∈ X, xi ∈ {0, 1}, i = 1, . . . , n,

where X is some set. Construct a branch-and-bound tree that starts with a sub-
problem where the integer constraints are relaxed, and proceeds with successive
restriction of the variables x1 , . . . , xn to the values 0 or 1.
(a) Show that the original integer-constrained problem is equivalent to a single
origin/single destination shortest path problem that involves the branch-
and-bound tree. Hint: As an example, for the traveling salesman problem,
nodes of the tree correspond to sequences (i1 , . . . , ik ) of distinct cities, and
arcs correspond to pairs of nodes (i1 , . . . , ik ) and (i1 , . . . , ik , ik+1 ).
(b) Modify the label correcting method of Section 2.5.2 so that it becomes
similar to the branch-and-bound method (see also the discussion in Section
2.5.2).

10.19

Use the branch-and-bound method to solve the capacitated spanning tree problem
of Fig. 10.5.
Sec. 10.6 Notes, Sources, and Exercises 535

10.20 (Simulated Annealing)

In the context of simulated annealing, assume that T is kept constant and let pxy
be the probability that when the current solution is x, the next solution sampled
is y. Consider the special case where pxy = pyx for all feasible solutions x and y,
and assume that the Markov chain defined by the probabilities pxy is irreducible,
in the sense that there is positive probability to go from any x to any y, with one
or more samples. Show that the steady-state probability of a solution x is

e−f (x)/T
πx = ,
C

where 
C= e−f (x)/T .
x∈F

Hint: This exercise assumes some basic knowledge of the theory of Markov chains.
Let qxy be the probability that y is the next solution if x is the current solution,
i.e.,  

− f (y)−f (x) /T
qxy = pxy e if f (y) > f (x),
pxy otherwise.

Show that for all x and y we have πy qyx = πx qxy , and that πy = π q .
x∈F x xy
This equality together with π = 1 is sufficient to show the result.
x∈F x

10.21 (Rollout Algorithms Based on Greedy Algorithms)

In the
context of the rollout algorithm, suppose that given a partial solution
P = xij | (i, j) ∈ S , we have an estimate c(P ) of the optimal cost over all
feasible solutions that are consistent

with P , in the sense that there exists a
complementary solution P = xij | (i, j) ∈ / S such that P ∪ P is feasible.
Consider a heuristic algorithm, which is greedy with respect to
c(P ), in the sense

that it starts from S = Ø, and given the partial solution P = xij | (i, j) ∈ S ,
it selects a
set of arcs T , 
forms the collection FT of all possible values of the arc
flows y = yij | (i, j) ∈ T , and finds

y = arg min c(Py+ ). (10.36)

y∈FT

where



Py+ = xij | (i, j) ∈ S , yij | (i, j) ∈ T .

It then augments P with the arc flows y thus obtained, and repeats up to ob-
taining a complete solution. Assume that the set of arcs T selected depends only
on P . Furthermore, the ties in the minimization of Eq. (10.36) are resolved in a
fixed manner that depends only on P . Show that the rollout algorithm that uses
the greedy algorithm as a base heuristic is sequentially consistent.
536 Network Problems with Integer Constraints Chap. 10

10.22 (Sequentially Improving Rollout Algorithm)

Consider a variant of the rollout algorithm that starts with the empty

set of arcs,
and maintains, in addition to the current partial solution P = xij | (i, j) ∈


S , a complementary solution P  = xij | (i, j) ∈ / S , and the corresponding
(complete) flow vector x = P ∪ P  . At the typical iteration, we select a subset T
of arcs that are not in

S, and we consider
 the collection FT of all possible values
of the arc flows y = yij | (i, j) ∈ T . Then, if

min H(Py+ ) < f (x ),

y∈FT


we augment the current partial solution {xij | (i, j) ∈ S with the arc flows


y = y ij | (i, j) ∈ T that attain the minimum above, and we set x equal to the
complete solution generated by the base heuristic starting from Py+ . Otherwise,

we augment the current partial solution to {xij | (i, j) ∈ S with the arc flows

 
xij | (i, j) ∈ T and we leave x unchanged. Prove that this rollout algorithm is
sequentially improving in the sense that the heuristic costs of the partial solutions
generated are monotonically nonincreasing.

10.23 (Scheduling Problems Viewed as Assignment Problems)

A machine can be used to perform a subset of N given tasks over T time periods.
At each time period t, only a subset A(t) of tasks can be performed. Each task
j has value vj (t) when performed at period t.
(a) Formulate the problem of finding the sequence of tasks of maximal total
value as an assignment problem. Hint: Assign time periods to tasks.
(b) Suppose that there are in addition some precedence constraints of the gen-
eral form: Task j must be performed before task j  can be performed.
Formulate the problem as an assignment problem with side constraints
and integer constraints. Give an example where the integer constraints are
essential.
(c) Repeat part (b) for the case where there are no precedence constraints, but
instead some of the tasks require more than one time period.

10.24 (Scheduling and the Interchange Argument)

In some scheduling problems it is useful to try to characterize a globally optimal

solution based on the fact that it is locally optimal with respect to the 2-OPT
heuristic. This is known as the interchange argument, and amounts to starting
with an optimal schedule and checking to see what happens when any two tasks
in the schedule are interchanged. As an example, suppose that we have N jobs
to process in sequential order with the ith job requiring a given time Ti for its
execution. If job i is completed at time t, the reward is αt Ri , where α is a given
discount factor with 0 < α < 1. The problem is to find a schedule that maximizes
Sec. 10.6 Notes, Sources, and Exercises 537

the total reward. Suppose that L = (i0 , . . . , ik−1 , i, j, ik+2 , . . . , iN −1 ) is an opti-

mal job schedule, and consider the schedule L = (i0 , . . . , ik−1 , j, i, ik+2 , . . . , iN −1 )
obtained by interchanging i and j. Let tk be the time of completion of job ik−1 .
Compare the rewards of the two schedules, and show that

αTi Ri αTj Rj
≥ .
1−α T i 1 − αTj
 
Conclude that scheduling jobs in order of decreasing αTi Ri / 1 − αTi is optimal.

10.25 (Weighted Shortest Processing Time First Rule)

We want to schedule N tasks, the ith of which requires Ti time units. Let ti
denote the time of completion of the ith task, i.e.,


ti = Ti + Tk .
tasks k
completed before i

Let wi denote a positive weight indicating the importance of early completion of

the ith task. Use an interchange argument (cf. Exercise 10.24) to show that in
N
order to minimize the total weighted completion time w t we must order
i=1 i i
the tasks in decreasing order of wi /Ti .

10.26

A busy professor has to complete N projects. Each project i has a deadline di

and the time it takes the professor to complete it is Ti . The professor can work
on only one project at a time and must complete it before moving on to a new
project. For a given order of completion of the projects, denote by ti the time of
completion of project i, i.e.,


ti = Ti + Tk .
projects k
completed before i

The professor wants to order the projects so as to minimize the maximum tardi-
ness, given by
max max[0, ti − di ].
i∈{1,...,N }

Use an interchange argument (cf. Exercise 10.24) to show that it is optimal to

complete the projects in the order of their deadlines (do the project with the
closest deadline first).
538 Network Problems with Integer Constraints Chap. 10

10.27 (Hardy’s Theorem)

Let {a1 , . . . , an } and {b1 , . . . , bn } be monotonically nondecreasing sequences of

numbers. Let us associate with each i = 1, . . . , n a distinct index ji , and consider
n
the expression i=1
a i b ji Use an interchange argument (cf. Exercise 10.24) to
.
show that this expression is maximized when ji = i for all i, and is minimized
when ji = n − i + 1 for all i.

10.28 (The Quiz Problem)

Consider a quiz contest where a person is given a list of N questions and can
answer these questions in any order he chooses. Question i will be answered
correctly with probability pi , independently of earlier answers, and the person will
then receive a reward Ri . At the first incorrect answer, the quiz terminates and
the person is allowed to keep his previous rewards. The problem is to maximize
the expected reward by choosing optimally the ordering of the questions.
(a) Show that to maximize the expected reward, questions should be answered
in decreasing order of pi Ri /(1 − pi ). Hint: Use an interchange argument
(cf. Exercise 10.24).
(b) Consider the variant of the problem where there is a maximum number
of questions that can be answered, which is smaller than the number of
questions that are available. Show that it is not necessarily optimal to
answer the questions in order of decreasing pi Ri /(1 − pi ). Hint: Try the
case where only one out of two available questions can be answered.
(c) Give a 2-OPT algorithm to solve the problem where the number of available
questions is one more than the maximum number of questions that can be
answered.

10.29 (Rollout Algorithm for the Quiz Problem)

Consider the quiz problem of Exercise 10.28 for the case where the maximum
number of questions that can be answered is less or equal to the number of
questions that are available. Consider the heuristic which answers questions in
decreasing order of pi Ri /(1 − pi ), and use it as a base heuristic in a rollout
algorithm. Show that the cost of the rollout algorithm is no worse than the cost
of the base heuristic. Hint: Prove sequential consistency of the base heuristic.

10.30

This exercise shows that nondifferentiabilities of the dual function given in Section
10.3, often tend to arise at the most interesting points and thus cannot be ignored.
Show that if there is a duality gap, then the dual function q is nondifferentiable
at every dual optimal solution. Hint: Assume that q has a unique subgradient at
a dual optimal solution µ∗ and derive a contradiction by showing that any vector
xµ∗ that minimizes L(x, µ∗ ) is primal optimal.
Sec. 10.6 Notes, Sources, and Exercises 539

10.31 (Enforced Separation in 3-Dimensional Assignment)

Consider the 3-dimensional assignment problem of Example 10.7 that involves a

set of jobs J, a set of machines M , and a set of workers W . We assume that
each of the sets J, M , and W contains n elements, and that the constraints are
equality constraints. Suppose that the problem is -separable, in the sense that
for some β jm and γ mw , and some  ≥ 0, we have

|β jm + γ mw − ajmw | ≤ , ∀ j ∈ J, m ∈ M, w ∈ W,
where ajmw is the value of the triplet (j, m, w).
(a) Show that if the problem is solved with ajmw replaced by β jm + γ mw , the
3-dimensional assignment obtained achieves the optimal cost of the original
problem within 2n.
(b) Suppose that we don’t know β jm and γ mw , and that we use the enforced
separation approach of Example 10.7. Thus, we first solve the jobs-to-
machines 2-dimensional assignment problem with values
bjm = max ajmw .
w∈W

Let jm be the job assigned to machine m, according to the solution of this

problem. We then solve the machines-to-workers 2-dimensional assignment
problem with values
cmw = ajm mw .
Let wm be the worker assigned to machine m, according to the solution of
this problem. Show that the 3-dimensional assignment {(jm , m, wm ) | m ∈
M } achieves the optimal value of the original problem within 4n.
(c) Show that the result of part (b) also holds when bjm is defined by
bjm = ajmwm ,
where wm is any worker, instead of bjm = maxw∈W ajmw .
(d) Show that the result of parts (b) and (c) also holds if J and W contain
more than n elements, and we have the inequality constraints
 
xjmw ≤ 1, ∀ j ∈ J,
m∈M w∈W

xjmw ≤ 1, ∀ w ∈ W,
j∈J m∈M

in place of equality constraints.

10.32 (Lagrangian Relaxation in Multidimensional Assignment)

Apply the Lagrangian relaxation method to the multidimensional assignment

problem of Example 10.7, in a way that requires the solution of 2-dimensional as-
signment problems. Derive the form of the corresponding subgradient algorithm.
540 Network Problems with Integer Constraints Chap. 10

10.33 (Separable Problems with Integer/Simplex Constraints)

Consider the problem


n

minimize fj (xj )
j=1


n

subject to xj ≤ A,
j=1

xj ∈ {0, 1, . . . , mj }, j = 1, . . . , n,

where A and m1 , . . . , mn are given positive integers, and each function fj is con-
vex over the interval [0, mj ]. Consider an iterative algorithm (due to Ibaraki and
Katoh [1988]) that starts at (0, . . . , 0) and maintains a feasible vector (x1 , . . . , xn ).
At the typical iteration, we consider the set of indices J = {j | xj < mj }. If
n
J is empty or j=1 j
x = A, the algorithm terminates. Otherwise, we find an
index j ∈ J that maximizes fj (xj ) − fj (xj + 1). If fj (xj ) − fj (xj + 1) ≤ 0, the
algorithm terminates. Otherwise, we increase xj by one unit, and go to the next
iteration. Show that upon termination, the algorithm yields an optimal solution.
Note: The book by Ibaraki and Katoh [1988] contains a lot of material on this
problem, and addresses the issues of efficient implementation.

10.34 (Constraint Relaxation and Lagrangian Relaxation)

The purpose of this exercise is to compare the lower bounds obtained by relaxing
integer constraints and by dualizing the side constraints. Consider the nonlinear
network optimization problem with a cost function f (x), the conservation of flow
constraints, and the additional constraint


x ∈ X = x | xij ∈ Xij , (i, j) ∈ A, gt (x) ≤ 0, t = 1, . . . , r ,

where Xij are given subsets of the real line and the functions gt are linear . We
assume that f is convex over the entire space of flow vectors x. We introduce a
Lagrange multiplier µt for each of the side constraints gt (x) ≤ 0, and we form
the corresponding Lagrangian function


r

L(x, µ) = f (x) + µt gt (x).

t=1

Let C denote the set of all x satisfying the conservation of flow constraints, let
f ∗ denote the optimal primal cost,

f∗ = inf f (x),
x∈C, xij ∈Xij , gt (x)≤0

and let q ∗ denote the optimal dual cost,

q ∗ = sup q(µ) = sup inf L(x, µ).

µ≥0 µ≥0 x∈C, xij ∈Xij
Sec. 10.6 Notes, Sources, and Exercises 541

Let X̂ ij denote the interval which is the convex hull of the set Xij , and denote
by fˆ the optimal cost of the problem, where each set Xij is replaced by X̂ ij ,
fˆ = inf f (x). (10.37)
x∈C, xij ∈X̂ ij , gt (x)≤0

Note that this is a convex problem even if Xij embodies integer constraints.
(a) Show that fˆ ≤ q ∗ ≤ f ∗ . Hint: Use Prop. 8.3 to show that problem (10.37)
has no duality gap and compare its dual cost with q ∗ .
(b) Assume that f is linear. Show that fˆ = q ∗ . Hint: The problem involved
in the definition of the dual function of problem (10.37) is a minimum cost
flow problem.
(c) Assume that C is a general polyhedron; that is, C is specified by a finite
number of linear equality and inequality constraints (rather than the con-
servation of flow constraints). Provide an example where f is linear and
we have fˆ < q ∗ .

10.35 (Duality Gap of the Knapsack Problem)

Given objects i = 1, . . . , n with positive weights wi and values vi , we want to

assemble a subset of the objects so that the sum of the weights of the subset does
not exceed a given T > 0, and the sum of the values of the subset is maximized.
This is the knapsack problem, which is a special case of a generalized assignment
problem (see Example 8.7). The problem can be written as

n

maximize vi xi
i=1
 n

subject to wi xi ≤ T, xi ∈ {0, 1}, i = 1, . . . , n.

i=1

(a) Let f ∗ and q ∗ be the optimal primal and dual costs, respectively. Show
that
0 ≤ q ∗ − f ∗ ≤ max vi .
i=1,...,n

(b) Consider the problem where T is multiplied by a positive integer k and

each object is replaced by k replicas of itself, while the object weights and
values stay the same. Let f ∗ (k) and q ∗ (k) be the corresponding primal and
dual costs. Show that
q ∗ (k) − f ∗ (k) 1 maxi=1,...,n vi
≤ ,
f ∗ (k) k f∗
so that the relative value of the duality gap tends to 0 as k → ∞. Note:
This exercise illustrates a generic property of many separable problems with
integer constraints: as the number of variables increases, the duality gap
decreases in relative terms (see Bertsekas [1982], Section 5.5, or Bertsekas
[1995b], Section 5.1, for an analysis and a geometrical interpretation of this
phenomenon).
542 Network Problems with Integer Constraints Chap. 10

10.36 (Convergence of the Subgradient Method)

Consider the subgradient method µk+1 = [µk + sk g k ]+ , where the stepsize is

given by
q ∗ − q(µk )
sk =
g k 2
and q ∗ is the optimal dual cost (this stepsize requires knowledge of q ∗ , which
is very restrictive, but the following Exercise 10.37 removes this restriction).
Assume that there exists at least one optimal dual solution.
(a) Use Eq. (10.16) to show that {µk } is bounded.
(b) Use the fact that {g k } is bounded (since the dual function is piecewise
linear), and Eq. (10.16) to show that q(µk ) → q ∗ .

10.37 (A Convergent Variation of the Subgradient Method)

This exercise provides a convergence result for a common variation of the subgra-
dient method (the result is due to Brannlund [1993]; see also Goffin and Kiwiel
[1996]). Consider the iteration µk+1 = [µk + sk g k ]+ , where

q̃ − q(µk )
sk = .
g k 2

(a) Suppose that q̃ is an underestimate of the optimal dual cost q ∗ such that
q(µk ) < q̃ ≤ q ∗ . [Here q̃ is fixed and the algorithm stops at µk if q(µk ) ≥ q̃.]
Use the fact that {g k } is bounded to show that either for some k̄
we have 
q(µk̄ ) ≥ q̃ or else q(µk ) → q̃. Hint: Consider the function min q(µ), q̃
and use the results of Exercise 10.36.
(b) Suppose that q̃ is an overestimate of the optimal dual cost, that is, q̃ > q ∗ .
Use the fact that {g k } is bounded to show that the length of the path
traveled by the method is infinite, that is,

∞ 
∞
q̃ − q(µk )
sk g k  = = ∞.
g k 
k=0 k=0

(c) Let δ 0 and B be two positive scalars. Consider the following version of
the subgradient method. Given µk , apply successive subgradient iterations
with q̃ = q(µk ) + δ k in the stepsize formula in place of q(µ∗ ), until one of
the following two occurs:
(1) The dual cost exceeds q(µk ) + δ k /2.
(2) The length of the path traveled starting from µk exceeds B.

Then set µk+1 to the iterate with highest dual cost thus far. Furthermore,
in case (1), set δ k+1 = δ k , while in case (2), set δ k+1 = δ k /2. Use the fact
that {g k } is bounded to show that q(µk ) → q ∗ .
10.38 (Convergence Rate of the Subgradient Method)

Consider the subgradient method of Exercise 10.36, and let µ∗ be an optimal

dual solution.
(a) Show that √  
lim inf k q(µ∗ ) − q(µk ) = 0.
k→∞

∞  2
Hint: Use Eq. (10.16) to show that q(µ∗ ) − q(µk ) < ∞. Assume
√  ∗
 k=0
that k q(µ ) − q(µ ) ≥  for some  > 0 and arbitrarily large k, and
k

reach a contradiction.
(b) Assume that for some a > 0 and all k, we have q(µ∗ ) − q(µk ) ≥ aµ∗ − µk .
Use Eq. (10.16) to show that for all k we have

µk+1 − µ∗  ≤ rµk − µ∗ ,
$
where r = 1 − a2 /b2 and b is an upper bound on g k .

10.39

Consider the cutting plane method.

(a) Give an example where the generated sequence q(µk ) is not monotonically
nondecreasing.
(b) Give an example where, at the kth iteration, the method finds an optimal
dual solution µk but does not terminate because the criterion q(µk ) =
Qk (µk ) is not satisfied.

10.40 (Computational Rollout Problem)

Consider the rollout algorithm for the traveling salesman problem using as base
heuristic the nearest neighbor method, whereby we start from some simple path
and at each iteration, we add a node that does not close a cycle and minimizes
the cost of the enlarged path (see the paragraph following the description of
the rollout algorithm iteration in Section 10.5). Write a computer program to
apply this algorithm to the problem involving Hamilton’s 20-node graph (Exercise
1.35) for the case where all arcs have randomly chosen costs from the range
[0, 1]. For node pairs for which there is no arc, introduce an artificial arc with
cost randomly chosen from the range [100, 101]. Compare the performances of
the rollout algorithm and the nearest neighbor heuristic, and compile relevant
statistics by running a suitable large collection of randomly generated problem
instances. Verify that the rollout algorithm performs at least as well as the
nearest neighbor heuristic for each instance (since it is sequentially consistent).

555
 556 References

References

Aarts, E., and Lenstra, J. K., 1997. Local Search in Combinatorial Opti-
mization, Wiley, N. Y.
Ahuja, R. K., Magnanti, T. L., and Orlin, J. B., 1989. “Network Flows,”
in Handbooks in Operations Research and Management Science, Vol. 1,
Optimization, Nemhauser, G. L., Rinnooy-Kan, A. H. G., and Todd M. J.
(eds.), North-Holland, Amsterdam, pp. 211-369.
Ahuja, R. K., Mehlhorn, K., Orlin, J. B., and Tarjan, R. E., 1990. “Faster
Algorithms for the Shortest Path Problem,” J. ACM, Vol. 37, 1990, pp.
213-223.
Ahuja, R. K., and Orlin, J. B., 1987. Private Communication.
Ahuja, R. K., and Orlin, J. B., 1989. “A Fast and Simple Algorithm for
the Maximum Flow Problem,” Operations Research, Vol. 37, pp. 748-759.
Amini, M. M., 1994. “Vectorization of an Auction Algorithm for Linear
Cost Assignment Problem,” Comput. Ind. Eng., Vol. 26, pp. 141-149.
Arezki, Y., and Van Vliet, D., 1990. “A Full Analytical Implementation of
the PARTAN/Frank-Wolfe Algorithm for Equilibrium Assignment,” Trans-
portation Science, Vol. 24, pp. 58-62.
Assad, A. A., and Golden, B. L., 1995. “Arc Routing Methods and Appli-
cations,” Handbooks in OR and MS, Ball, M. O., Magnanti, T. L., Monma,
C. L., and Nemhauser, G. L., (eds.), Vol. 8, North-Holland, Amsterdam,
pp. 375-483.
Atkinson, D. S., and Vaidya, P. M., 1995. “A Cutting Plane Algorithm for
Convex Programming that Uses Analytic Centers,” Math. Programming,
Vol. 69, pp. 1-44.
Auchmuty, G., 1989. “Variational Principles for Variational Inequalities,”
Numer. Functional Analysis and Optimization, Vol. 10, pp. 863-874.
Auslender, A., 1976. Optimization: Methodes Numeriques, Mason, Paris.
Balas, E., Miller, D., Pekny, J., and Toth, P., 1991. “A Parallel Shortest
Path Algorithm for the Assignment Problem,” J. ACM, Vol. 38, pp. 985-
1004.
Balas, E., and Toth, P., 1985. “Branch and Bound Methods,” in The Trav-
eling Salesman Problem, Lawler, E., Lenstra, J. K., Rinnoy Kan, A. H. G.,
and Shmoys, D. B. (eds.), Wiley, N. Y., pp. 361-401.
References 557

Balinski, M. L., 1985. “Signature Methods for the Assignment Problem,”

Operations Research, Vol. 33, pp. 527-537.
Balinski, M. L., 1986. “A Competitive (Dual) Simplex Method for the
Assignment Problem,” Math. Programming, Vol. 34, pp. 125-141.
Ball, M. O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L., 1995a.
Network Models, Handbooks in OR and MS, Vol. 7, North-Holland, Ams-
terdam.
Ball, M. O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L., 1995b.
Network Routing, Handbooks in OR and MS, Vol. 8, North-Holland, Am-
sterdam.
Bar-Shalom, Y., and Fortman, T. E., 1988. Tracking and Data Association,
Academic Press, N. Y.
Barnhart, C., Hane, C. H., and Vance, P. H., 1997. “Integer Multicommod-
ity Flow Problems,” in Network Optimization, Pardalos, P. M., Hearn, D.
W., and Hager, W. W. (eds.), Springer-Verlag, N. Y., pp. 17-31.
Barr, R., Glover, F., and Klingman, D., 1977. “The Alternating Basis
Algorithm for Assignment Problems,” Math. Programming, Vol. 13, pp.
1-13.
Barr, R., Glover, F., and Klingman, D., 1978. “Generalized Alternating
Path Algorithm for Transportation Problems,” European J. of Operations
Research, Vol. 2, pp. 137-144.
Barr, R., Glover, F., and Klingman, D., 1979. “Enhancement of Spanning
Tree Labeling Procedures for Network Optimization,” INFOR, Vol. 17, pp.
16-34.
Barr, R., and Hickman, B. L., 1994. “Parallel Simplex for Large Pure
Network Problems - Computational Testing and Sources of Speedup,” Op-
erations Research, Vol. 42, pp. 65-80.
Bazaraa, M. S., Jarvis, J. J., and Sherali, H. D., 1990. Linear Programming
and Network Flows (2nd Ed.), Wiley, N. Y.
Bazaraa, M. S., Sherali, H. D., and Shetty, C. M., 1993. Nonlinear Pro-
gramming Theory and Algorithms (2nd Ed.), Wiley, N. Y.
Bell, G. J., and Lamar, B. W., 1997. “Solution Methods for Nonconvex Net-
work Flow Problems,” in Network Optimization, Pardalos, P. M., Hearn,
D. W., and Hager, W. W. (eds.), Lecture Notes in Economics and Mathe-
matical Systems, Springer-Verlag, N. Y., pp. 32-50.
Bellman, R., 1957. Dynamic Programming, Princeton Univ. Press, Prince-
ton, N. J.
Benders, J. F., 1962. “Partitioning Procedures for Solving Mixed Variables
Programming Problems,” Numer. Math., Vol. 4, pp. 238-252.
558 References

Beraldi, P., and Guerriero, F., 1997. “A Parallel Asynchronous Implemen-

tation of the Epsilon-Relaxation Method for the Linear Minimum Cost
Flow Problem,” Parallel Computing, Vol. 23, pp. 1021-1044.
Beraldi, P., Guerriero, F., and Musmanno, R., 1996. “Parallel Algorithms
for Solving the Convex Minimum Cost Flow Problem,” Tech. Report PAR-
COLAB No. 8/96, Dept. of Electronics, Informatics, and Systems, Univ.
of Calabria.
Beraldi, P., Guerriero, F., and Musmanno, R., 1997. “Efficient Parallel Al-
gorithms for the Minimum Cost Flow Problem,” J. of Optimization Theory
and Applications, Vol. 95, pp. 501-530.
Berge, C., 1962. The Theory of Graphs and its Applications, Wiley, N. Y.
Berge, C., and Ghouila-Houri, A., 1962. Programming, Games, and Trans-
portation Networks, Wiley, N. Y.
Bertsekas, D. P., 1975a. “Nondifferentiable Optimization via Approxima-
tion,” Math. Programming Studies, Vol. 3, North-Holland, Amsterdam,
pp. 1-25.
Bertsekas, D. P., 1975b. “Necessary and Sufficient Conditions for a Penalty
Method to be Exact,” Math. Programming, Vol. 9, pp. 87-99.
Bertsekas, D. P., 1979a. “A Distributed Algorithm for the Assignment
Problem,” Lab. for Information and Decision Systems Working Paper,
M.I.T., Cambridge, MA.
Bertsekas, D. P., 1979b. “Algorithms for Nonlinear Multicommodity Net-
work Flow Problems,” in International Symposium on Systems Optimiza-
tion and Analysis, Bensoussan, A., and Lions, J. L. (eds.), Springer-Verlag,
N. Y., pp. 210-224.
Bertsekas, D. P., 1980. “A Class of Optimal Routing Algorithms for Com-
munication Networks,” Proc. of the Fifth International Conference on Com-
puter Communication, Atlanta, Ga., pp. 71-76.
Bertsekas, D. P., 1981. “A New Algorithm for the Assignment Problem,”
Math. Programming, Vol. 21, pp. 152-171.
Bertsekas, D. P., 1982. Constrained Optimization and Lagrange Multiplier
Methods, Academic Press, N. Y. (republished in 1996 by Athena Scientific,
Belmont, MA).
Bertsekas, D. P., 1985. “A Unified Framework for Minimum Cost Network
Flow Problems,” Math. Programming, Vol. 32, pp. 125-145.
Bertsekas, D. P., 1986a. “Distributed Asynchronous Relaxation Methods
for Linear Network Flow Problems,” Lab. for Information and Decision
Systems Report P-1606, M.I.T., Cambridge, MA.
Bertsekas, D. P., 1986b. “Distributed Relaxation Methods for Linear Net-
References 559

work Flow Problems,” Proceedings of 25th IEEE Conference on Decision

and Control, Athens, Greece, pp. 2101-2106.
Bertsekas, D. P., 1988. “The Auction Algorithm: A Distributed Relaxation
Method for the Assignment Problem,” Annals of Operations Research, Vol.
14, pp. 105-123.
Bertsekas, D. P., 1990. “The Auction Algorithm for Assignment and Other
Network Flow Problems: A Tutorial,” Interfaces, Vol. 20, pp. 133-149.
Bertsekas, D. P., 1991a. Linear Network Optimization: Algorithms and
Codes, MIT Press, Cambridge, MA.
Bertsekas, D. P., 1991b. “An Auction Algorithm for Shortest Paths,” SIAM
J. on Optimization, Vol. 1, pp. 425-447.
Bertsekas, D. P., 1992a. “Auction Algorithms for Network Flow Problems:
A Tutorial Introduction,” Computational Optimization and Applications,
Vol. 1, pp. 7-66.
Bertsekas, D. P., 1992b. “Modified Auction Algorithms for Shortest Paths,”
Lab. for Information and Decision Systems Report P-2150, M.I.T., Cam-
bridge, MA.
Bertsekas, D. P., 1992c. “An Auction Sequential Shortest Path Algorithm
for the Minimum Cost Network Flow Problem,” Lab. for Information and
Decision Systems Report P-2146, M.I.T.
Bertsekas, D. P., 1993a. “A Simple and Fast Label Correcting Algorithm
for Shortest Paths,” Networks, Vol. 23, pp. 703-709.
Bertsekas, D. P., 1993b. “Mathematical Equivalence of the Auction Algo-
rithm for Assignment and the
-Relaxation (Preflow-Push) Method for Min
Cost Flow,” in Large Scale Optimization: State of the Art, Hager, W. W.,
Hearn, D. W., and Pardalos, P. M. (eds.), Kluwer, Boston, pp. 27-46.
Bertsekas, D. P., 1995a. Dynamic Programming and Optimal Control, Vols.
I and II, Athena Scientific, Belmont, MA.
Bertsekas, D. P., 1995b. Nonlinear Programming, Athena Scientific, Bel-
mont, MA.
Bertsekas, D. P., 1995c. “An Auction Algorithm for the Max-Flow Prob-
lem,” J. of Optimization Theory and Applications, Vol. 87, pp. 69-101.
Bertsekas, D. P., 1996. “Thevenin Decomposition and Network Optimiza-
tion,” J. of Optimization Theory and Applications, Vol. 89, pp. 1-15.
Bertsekas, D. P., and Castañon, D. A., 1989. “The Auction Algorithm for
Transportation Problems,” Annals of Operations Research, Vol. 20, pp.
67-96.
Bertsekas, D. P., and Castañon, D. A., 1991. “Parallel Synchronous and
560 References

Asynchronous Implementations of the Auction Algorithm,” Parallel Com-

puting, Vol. 17, pp. 707-732.
Bertsekas, D. P., and Castañon, D. A., 1992. “A Forward/Reverse Auction
Algorithm for Asymmetric Assignment Problems,” Computational Opti-
mization and Applications, Vol. 1, pp. 277-297.
Bertsekas, D. P., and Castañon, D. A., 1993a. “Asynchronous Hungarian
Methods for the Assignment Problem,” ORSA J. on Computing, Vol. 5,
pp. 261-274.
Bertsekas, D. P., and Castañon, D. A., 1993b. “Parallel Primal-Dual Meth-
ods for the Minimum Cost Flow Problem,” Computational Optimization
and Applications, Vol. 2, pp. 317-336.
Bertsekas, D. P., and Castañon, D. A., 1993c. “A Generic Auction Al-
gorithm for the Minimum Cost Network Flow Problem,” Computational
Optimization and Applications, Vol. 2, pp. 229-260.
Bertsekas, D. P., and Castañon, D. A., 1998. “Solving Stochastic Scheduling
Problems Using Rollout Algorithms,” Lab. for Information and Decision
Systems Report P-12413, M.I.T., Cambridge, MA.
Bertsekas, D. P., Castañon, D. A., Eckstein, J., and Zenios, S., 1995. “Par-
allel Computing in Network Optimization,” Handbooks in OR and MS,
Ball, M. O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L. (eds.),
Vol. 7, North-Holland, Amsterdam, pp. 331-399.
Bertsekas, D. P., Castañon, D. A., and Tsaknakis, H., 1993. “Reverse Auc-
tion and the Solution of Inequality Constrained Assignment Problems,”
SIAM J. on Optimization, Vol. 3, pp. 268-299.
Bertsekas, D. P., and El Baz, D., 1987. “Distributed Asynchronous Relax-
ation Methods for Convex Network Flow Problems,” SIAM J. on Control
and Optimization, Vol. 25, pp. 74-85.
Bertsekas, D. P., and Eckstein, J., 1987. “Distributed Asynchronous Re-
laxation Methods for Linear Network Flow Problems,” Proc. of IFAC ’87,
Munich, Germany.
Bertsekas, D. P., and Eckstein, J., 1988. “Dual Coordinate Step Methods
for Linear Network Flow Problems,” Math. Programming, Series B, Vol.
42, pp. 203-243.
Bertsekas, D. P., and Gafni, E. M., 1982. “Projection Methods for Varia-
tional Inequalities with Application to the Traffic Assignment Problem,”
Math. Progr. Studies, Vol. 17, North-Holland, Amsterdam, pp. 139-159.
Bertsekas, D. P., and Gafni, E. M., 1983. “Projected Newton Methods and
Optimization of Multicommodity Flows,” IEEE Trans. on Auto. Control,
Vol. 28, pp. 1090-1096.
References 561

Bertsekas, D. P., Gafni, E. M., and Gallager, R. G., 1984. “Second Deriva-
tive Algorithms for Minimum Delay Distributed Routing in Networks,”
IEEE Trans. on Communications, Vol. 32, pp. 911-919.
Bertsekas, D. P., and Gallager, R. G., 1992. Data Networks, (2nd Ed.),
Prentice-Hall, Englewood Cliffs, N. J.
Bertsekas, D. P., Guerriero, F., and Musmanno, R., 1996. “Parallel Asyn-
chronous Label Correcting Methods for Shortest Paths,” J. of Optimization
Theory and Applications, Vol. 88, pp. 297-320.
Bertsekas, D. P., Hosein, P., and Tseng, P., 1987. “Relaxation Methods for
Network Flow Problems with Convex Arc Costs,” SIAM J. on Control and
Optimization, Vol. 25, pp. 1219-1243.
Bertsekas, D. P, and Mitter, S. K., 1971. “Steepest Descent for Optimiza-
tion Problems with Nondifferentiable Cost Functionals,” Proc. 5th Annual
Princeton Confer. Inform. Sci. Systems, Princeton, N. J., pp. 347-351.
Bertsekas, D. P., and Mitter, S. K., 1973. “Descent Numerical Methods for
Optimization Problems with Nondifferentiable Cost Functions,” SIAM J.
on Control, Vol. 11, pp. 637-652.
Bertsekas, D. P., Pallottino, S., and Scutellà, M. G., 1995. “Polynomial
Auction Algorithms for Shortest Paths,” Computational Optimization and
Applications, Vol. 4, pp. 99-125.
Bertsekas, D. P., Polymenakos, L. C., and Tseng, P., 1997a. “An
-Relaxati-
on Method for Separable Convex Cost Network Flow Problems,” SIAM J.
on Optimization, Vol. 7, pp. 853-870.
Bertsekas, D. P., Polymenakos, L. C., and Tseng, P., 1997b. “Epsilon-
Relaxation and Auction Methods for Separable Convex Cost Network Flow
Problems,” in Network Optimization, Pardalos, P. M., Hearn, D. W., and
Hager, W. W. (eds.), Lecture Notes in Economics and Mathematical Sys-
tems, Springer-Verlag, N. Y., pp. 103-126.
Bertsekas, D. P., and Tseng, P., 1988a. “Relaxation Methods for Minimum
Cost Ordinary and Generalized Network Flow Problems,” Operations Re-
search, Vol. 36, pp. 93-114.
Bertsekas, D. P., and Tseng, P., 1988b. “RELAX: A Computer Code for
Minimum Cost Network Flow Problems,” Annals of Operations Research,
Vol. 13, pp. 127-190.
Bertsekas, D. P., and Tseng, P., 1990. “RELAXT-III: A New and Improved
Version of the RELAX Code,” Lab. for Information and Decision Systems
Report P-1990, M.I.T., Cambridge, MA.
Bertsekas, D. P., and Tseng, P., 1994. “RELAX-IV: A Faster Version of
the RELAX Code for Solving Minimum Cost Flow Problems,” Laboratory
for Information and Decision Systems Report P-2276, M.I.T., Cambridge,
562 References

MA.
Bertsekas, D. P., and Tsitsiklis, J. N., 1989. Parallel and Distributed Com-
putation: Numerical Methods, Prentice-Hall, Englewood Cliffs, N. J. (re-
published in 1997 by Athena Scientific, Belmont, MA).
Bertsekas, D. P., and Tsitsiklis, J. N., 1996. Neuro-Dynamic Programming,
Athena Scientific, Belmont, MA.
Bertsekas, D. P., Tsitsiklis, J. N., and Wu, C., 1997. “Rollout Algorithms
for Combinatorial Optimization,” Heuristics, Vol. 3, pp. 245-262.
Bertsimas, D., and Tsitsiklis, J. N., 1993. “Simulated Annealing,” Stat.
Sci., Vol. 8, pp. 10-15.
Bertsimas, D., and Tsitsiklis, J. N., 1997. Introduction to Linear Optimiza-
tion, Athena Scientific, Belmont, MA.
Birkhoff, G., and Diaz, J. B., 1956. “Nonlinear Network Problems,” Quart.
Appl. Math., Vol. 13, pp. 431-444.
Bland, R. G., and Jensen, D. L., 1985. “On the Computational Behavior
of a Polynomial-Time Network Flow Algorithm,” Tech. Report 661, School
of Operations Research and Industrial Engineering, Cornell University.
Blackman, S. S., 1986. Multi-Target Tracking with Radar Applications,
Artech House, Dehdam, MA.
Bogart, K. P., 1990. Introductory Combinatorics, Harcourt Brace Jovano-
vich, Inc., New York, N. Y.
Bradley, G. H., Brown, G. G., and Graves, G. W., 1977. “Design and Imple-
mentation of Large-Scale Primal Transshipment Problems,” Management
Science, Vol. 24, pp. 1-38.
Brannlund, U., 1993. On Relaxation Methods for Nonsmooth Convex Opti-
mization, Doctoral Thesis, Royal Institute of Technology, Stockhorm, Swe-
den.
Brown, G. G., and McBride, R. D., 1984. “Solving Generalized Networks,”
Management Science, Vol. 30, pp. 1497-1523.
Burkard, R. E., 1990. “Special Cases of Traveling Salesman Problems and
Heuristics,” Acta Math. Appl. Sin., Vol. 6, pp. 273-288.
Busacker, R. G., and Gowen, P. J., 1961. “A Procedure for Determining a
Family of Minimal-Cost Network Flow Patterns,” O.R.O. Technical Report
No. 15, Operational Research Office, John Hopkins University, Baltimore,
MD.
Busacker, R. G., and Saaty, T. L., 1965. Finite Graphs and Networks: An
Introduction with Applications, McGraw-Hill, N. Y.
References 563

Cameron, P. J., 1994. Combinatorics: Topics, Techniques, Algorithms,

Cambridge Univ. Press, Cambridge, England.
Cantor, D. G., and Gerla, M., 1974. “Optimal Routing in a Packet Switched
Computer Network,” IEEE Trans. on Computers, Vol. 23, pp. 1062-1069.
Carpaneto, G., Martello, S., and Toth, P., 1988. “Algorithms and Codes
for the Assignment Problem,” Annals of Operations Research, Vol. 13, pp.
193-223.
Carraresi, P., and Sodini, C., 1986. “An Efficient Algorithm for the Bipar-
tite Matching Problem,” Eur. J. Operations Research, Vol. 23, pp. 86-93.
Castañon, D. A., 1990. “Efficient Algorithms for Finding the K Best Paths
Through a Trellis,” IEEE Trans. on Aerospace and Electronic Systems,
Vol. 26, pp. 405-410.
Castañon, D. A., 1993. “Reverse Auction Algorithms for Assignment Prob-
lems,” in Algorithms for Network Flows and Matching, Johnson, D. S., and
McGeoch, C. C. (eds.), American Math. Soc., Providence, RI, pp. 407-429.
Censor, Y., and Zenios, S. A., 1992. “The Proximal Minimization Algo-
rithm with D-Functions,” J. Opt. Theory and Appl., Vol. 73, pp. 451-464.
Censor, Y., and Zenios, S. A., 1997. Parallel Optimization: Theory, Algo-
rithms, and Applications, Oxford University Press, N. Y.
Cerny, V., 1985. “A Thermodynamical Approach to the Travelling Sales-
man Problem: An Efficient Simulation Algorithm,” J. Opt. Theory and
Applications, Vol. 45, pp. 41-51.
Cerulli, R., De Leone, R., and Piacente, G., 1994. “A Modified Auction
Algorithm for the Shortest Path Problem,” Optimization Methods and
Software, Vol. 4, pp. 209-224.
Cerulli, R., Festa, P., and Raiconi, G., 1997a. “Graph Collapsing in Short-
est Path Auction Algorithms,” Univ. of Salerno Tech. Report n. 6/97.
Cerulli, R., Festa, P., and Raiconi, G., 1997b. “An Efficient Auction Algo-
rithm for the Shortest Path Problem Using Virtual Source Concept,” Univ.
of Salerno Tech. Report n. 6/97.
Chajakis, E. D., and Zenios, S. A., 1991. “Synchronous and Asynchronous
Implementations of Relaxation Algorithms for Nonlinear Network Opti-
mization,” Parallel Computing, Vol. 17, pp. 873-894.
Chen, G., and Teboulle, M., 1993. “Convergence Analysis of a Proximal-
Like Minimization Algorithm Using Bregman Functions,” SIAM J. on Op-
timization, Vol. 3, pp. 538-543.
Chen, Z. L., and Powell, W. B., 1997. “A Note on Bertsekas’ Small-Label-
First Strategy,” Networks, Vol. 29, pp. 111-116.
564 References

Cheney, E. W., and Goldstein, A. A., 1959. “Newton’s Method for Convex
Programming and Tchebycheff Approximation,” Numer. Math., Vol. I, pp.
253-268.
Cheriyan, J., and Maheshwari, S. N., 1989. “Analysis of Preflow Push
Algorithms for Maximum Network Flow,” SIAM J. Computing, Vol. 18,
pp. 1057-1086.
Cherkasky, R. V., 1977. “Algorithm√for Construction of Maximum Flow in
Networks with Complexity of O(V 2 E) Operations,” Mathematical Meth-
ods of Solution of Economical Problems, Vol. 7, pp. 112-125.
Christofides, N., 1975. Graph Theory: An Algorithmic Approach, Aca-
demic Press, N. Y.
Chvatal, V., 1983. Linear Programming, W. H. Freeman and Co., N. Y.
Connors, D. P., and Kumar, P. R., 1989. “Simulated Annealing Type
Markov Chains and their Order Balance Equations,” SIAM J. on Control
and Optimization, Vol. 27, pp. 1440-1461.
Cook, W., Cunningham, W., Pulleyblank, W., and Schrijver, A., 1998.
Combinatorial Optimization, Wiley, N. Y.
Cornuejols, G., Fonlupt, J., and Naddef, D., 1985. “The Traveling Salesman
Problem on a Graph and Some Related Polyhedra,” Math. Programming,
Vol. 33, pp. 1-27.
Cottle, R. W., and Pang, J. S., 1982. “On the Convergence of a Block
Successive Over-Relaxation Method for a Class of Linear Complementarity
Problems,” Math. Progr. Studies, Vol. 17, pp. 126-138.
Croes, G. A., 1958. “A Method for Solving Traveling Salesman Problems,”
Operations Research, Vol. 6, pp. 791-812.
Cunningham, W. H., 1976. “A Network Simplex Method,” Math. Program-
ming, Vol. 4, pp. 105-116.
Cunningham, W. H., 1979. “Theoretical Properties of the Network Simplex
Method,” Math. of Operations Research, Vol. 11, pp. 196-208.
Dafermos, S., 1980. “Traffic Equilibrium and Variational Inequalities,”
Transportation Science, Vol. 14, pp. 42-54.
Dafermos, S., 1982. “Relaxation Algorithms for the General Asymmetric
Traffic Equilibrium Problem,” Transportation Science, Vol. 16, pp. 231-240.
Dafermos, S., and Sparrow, F. T., 1969. “The Traffic Assignment Problem
for a General Network,” J. Res. Nat. Bureau of Standards, Vol. 73B, pp.
91-118.
Dantzig, G. B., 1951. “Application of the Simplex Method to a Transporta-
tion Problem,” in Activity Analysis of Production and Allocation, T. C.
References 565

Koopmans (ed.), Wiley, N. Y., pp. 359-373.

Dantzig, G. B., 1960. “On the Shortest Route Problem Through a Net-
work,” Management Science, Vol. 6, pp. 187-190.
Dantzig, G. B., 1963. Linear Programming and Extensions, Princeton Univ.
Press, Princeton, N. J.
Dantzig, G. B., 1967. “All Shortest Routes in a Graph,” in Theory of
Graphs, P. Rosenthier (ed.), Gordan and Breach, N. Y., pp. 92-92.
Dantzig, G. B., and Fulkerson, D. R., 1956. “On the Max-Flow Min-Cut
Theorem of Networks,” in Linear Inequalities and Related Systems, Kuhn,
H. W., and Tucker, A. W. (eds.), Annals of Mathematics Study 38, Prince-
ton Univ. Press, pp. 215-221.
Dantzig, G. B., and Wolfe, P., 1960. “Decomposition Principle for Linear
Programs,” Operations Research, Vol. 8, pp. 101-111.
Dantzig, G. B., Fulkerson, D. R., and Johnson, S. M., 1954. “Solution of
a Large-Scale Traveling-Salesman Problem,” Operations Research, Vol. 2,
pp. 393-410.
De Leone, R., Meyer, R. R., and Zakarian, A., 1995. “An
-Relaxation
Algorithm for Convex Network Flow Problems,” Computer Sciences De-
partment Technical Report, University of Wisconsin, Madison, WI.
Dembo, R. S., 1987. “A Primal Truncated Newton Algorithm for Large-
Scale Unconstrained Optimization,” Math. Programming Studies, Vol. 31,
pp. 43-72.
Dembo, R. S., and Klincewicz, J. G., 1981. “A Scaled Reduced Gradi-
ent Algorithm for Network Flow Problems with Convex Separable Costs,”
Math. Programming Studies, Vol. 15, pp. 125-147.
Dembo, R. S., and Tulowitzki, U., 1988. “Computing Equilibria on Large
Multicommodity Networks: An Application of Truncated Quadratic Pro-
gramming Algorithms,” Networks, Vol. 18, pp. 273-284.
Denardo, E. V., and Fox, B. L., 1979. “Shortest-Route Methods: 1. Reach-
ing, Pruning and Buckets,” Operations Research, Vol. 27, pp. 161-186.
Dennis, J. B., 1959. Mathematical Programming and Electical Circuits,
Technology Press of M.I.T., Cambridge, MA.
Deo, N., and Kumar, N., 1997. “Computation of Constrained Spanning
Trees: A Unified Approach,” in Network Optimization, Pardalos, P. M.,
Hearn, D. W., and Hager, W. W. (eds.), Lecture Notes in Economics and
Mathematical Systems, Springer-Verlag, N. Y., pp. 194-220.
Deo, N., and Pang, C., 1984. “Shortest Path Algorithms: Taxonomy and
Annotation,” Networks, Vol. 14, pp. 275-323.
566 References

Derigs, U., 1985. “The Shortest Augmenting Path Method for Solving As-
signment Problems – Motivation and Computational Experience,” Annals
of Operations Research, Vol. 4, pp. 57-102.
Derigs, U., and Meier, W., 1989. “Implementing Goldberg’s Max-Flow Al-
gorithm – A Computational Investigation,” Zeitschrif fur Operations Re-
search, Vol. 33, pp. 383-403.
Desrosiers, J., Dumas, Y., Solomon, M. M., and Soumis, F., 1995. “Time
Constrained Routing and Scheduling,” Handbooks in OR and MS, Ball,
M. O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol.
8, North-Holland, Amsterdam, pp. 35-139.
Dial, R. B., 1969. “Algorithm 360: Shortest Path Forest with Topological
Ordering,” Comm. ACM, Vol. 12, pp. 632-633.
Dial, R., Glover, F., Karney, D., and Klingman, D., 1979. “A Compu-
tational Analysis of Alternative Algorithms and Labeling Techniques for
Finding Shortest Path Trees,” Networks, Vol. 9, pp. 215-248.
Dijkstra, E., 1959. “A Note on Two Problems in Connexion with Graphs,”
Numer. Math., Vol. 1, pp. 269-271.
Dinic, E. A., 1970. “Algorithm for Solution of a Problem of Maximum Flow
in Networks with Power Estimation,” Soviet Math. Doklady, Vol. 11, pp.
1277-1280.
Dreyfus, S. E., 1969. “An Appraisal of Some Shortest-Path Algorithms,”
Operations Research, Vol. 17, pp. 395-412.
Duffin, R. J., 1947. “Nonlinear Networks. IIa,” Bull. Amer. Math. Soc.,
Vol. 53, pp. 963-971.
Eastman, W. L., 1958. Linear Programming with Pattern Constraints,
Ph.D. Thesis, Harvard University, Cambridge, MA.
Eckstein, J., 1994. “Nonlinear Proximal Point Algorithms Using Bregman
Functions, with Applications to Convex Programming,” Math. of Opera-
tions Research, Vol. 18, pp. 202-226.
Edmonds, J., 1965. “Paths, Trees, and Flowers,” Canadian J. of Math.,
Vol. 17, pp. 449-467.
Edmonds, J., and Johnson, E. L., 1973. “Matching, Euler Tours, and the
Chinese Postman,” Math. Programming, Vol. 5, pp. 88-124.
Edmonds, J., and Karp, R. M., 1972. “Theoretical Improvements in Al-
gorithmic Efficiency for Network Flow Problems,” J. ACM, Vol. 19, pp.
248-264.
Eiselt, H. A., Gendreau, M., and Laporte, G., 1995a. “Arc Routing Prob-
lems, Part 1: The Chinese Postman Problem,” Operations Research, Vol.
43, pp. 231-242.
References 567

Eiselt, H. A., Gendreau, M., and Laporte, G., 1995b. “Arc Routing Prob-
lems, Part 2: The Rural Postman Problem,” Operations Research, Vol. 43,
pp. 399-414.
Elias, P., Feinstein, A., and Shannon, C. E., 1956. “Note on Maximum Flow
Through a Network,” IRE Trans. Info. Theory, Vol. IT-2, pp. 117-119.
Egervary, J., 1931. “Matrixok Kombinatoricus Tulajonsagairol,” Mat. Es
Fiz. Lapok, Vol. 38, pp. 16-28.
El Baz, D., 1989. “A Computational Experience with Distributed Asyn-
chronous Iterative Methods for Convex Network Flow Problems,” Proc.
of the 28th IEEE Conference on Decision and Control, Tampa, Fl., pp.
590-591.
El Baz, D., 1996. “Asynchronous Gradient Algorithms for a Class of Con-
vex Separable Network Flow Problems,” Computational Optimization and
Applications, Vol. 5, pp. 187-205.
El Baz, D., Spiteri, P., Miellou, J. C., and Gazen, D., 1996. “Asynchronous
Iterative Algorithms with Flexible Communication for Nonlinear Network
Flow Problems,” J. of Parallel and Distributed Computing, Vol. 38, pp.
1-15.
Elam, J., Glover, F., and Klingman, D., 1979. “A Strongly Convergent
Primal Simplex Algorithm for Generalized Networks,” Math. of Operations
Research, Vol. 4, pp. 39-59.
Elmaghraby, S. E., 1978. Activity Networks: Project Planning and Control
by Network Models, Wiley, N. Y.
Elzinga, J., and Moore, T. G., 1975. “A Central Cutting Plane Algorithm
for the Convex Programming Problem,” Math. Programming, Vol. 8, pp.
134-145.
Engquist, M., 1982. “A Successive Shortest Path Algorithm for the Assign-
ment Problem,” INFOR, Vol. 20, pp. 370-384.
Ephremides, A., 1986. “The Routing Problem in Computer Networks,”
in Communication and Networks, Blake, I. F., and Poor, H. V. (eds.),
Springer-Verlag, N. Y., pp. 299-325.
Ephremides, A., and Verdu, S., 1989. “Control and Optimization Methods
in Communication Network Problems,” IEEE Trans. on Automatic Con-
trol, Vol. 34, pp. 930-942.
Esau, L. R., and Williams, K. C., 1966. “On Teleprocessing System Design.
A Method for Approximating the Optimal Network,” IBM System J., Vol.
5, pp. 142-147.
Escudero, L. F., 1985. “Performance Evaluation of Independent Superbasic
Sets on Nonlinear Replicated Networks,” Eur. J. Operations Research, Vol.
568 References

23, pp. 343-355.

Everett, H., 1963. “Generalized Lagrange Multiplier Method for Solving
Problems of Optimal Allocation of Resources,” Operations Research, Vol.
11, pp. 399-417.
Falcone, M., 1987. “A Numerical Approach to the Infinite Horizon Problem
of Deterministic Control Theory,” Appl. Math. Opt., Vol. 15, pp. 1-13.
Federgruen, A., and Simchi-Levi, D., 1995. “Analysis of Vehicle and Invento-
ry-Routing Problems,” Handbooks in OR and MS, Ball, M. O., Magnanti,
T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 8, North-Holland,
Amsterdam, pp. 297-373.
Ferris, M. C., 1991. “Finite Termination of the Proximal Point Algorithm,”
Math. Programming, Vol. 50, pp. 359-366.
Fisher, M., 1995. “Vehicle Routing,” Handbooks in OR and MS, Ball, M.
O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 8,
North-Holland, Amsterdam, pp. 1-33.
Florian, M., Guélat, J., and Spiess, H., 1987. “An Efficient Implementation
of the “PARTAN” Variant of the Linear Approximation Method for the
Network Equilibrium Problem,” Networks, Vol. 17, pp. 319-339.
Florian, M. S., and Hearn, D., 1995. “Network Equilibrium Models and
Algorithms,” Handbooks in OR and MS, Ball, M. O., Magnanti, T. L.,
Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 8, North-Holland, Ams-
terdam, pp. 485-550.
Florian, M. S., and Nguyen, S., 1974. “A Method for Computing Network
Equilibrium with Elastic Demands,” Transportation Science, Vol. 8, pp.
321-332.
Florian, M. S., and Nguyen, S., 1976. “An Application and Validation of
Equilibrium Trip Assignment Methods,” Transportation Science, Vol. 10,
pp. 374-390.
Florian, M. S., Nguyen, S., and Pallottino, S., 1981. “A Dual Simplex
Algorithm for Finding All Shortest Paths,” Networks, Vol. 11, pp. 367-
378.
Floudas, C. A., 1995. Nonlinear and Mixed-Integer Optimization: Funda-
mentals and Applications, Oxford University Press, N. Y.
Floyd, R. W., 1962. “Algorithm 97: Shortest Path,” Comm. ACM, Vol. 5,
pp. 345.
Ford, L. R., Jr., 1956. “Network Flow Theory,” Report P-923, The Rand
Corporation, Santa Monica, CA.
Ford, L. R., Jr., and Fulkerson, D. R., 1956a. “Solving the Transportation
Problem,” Management Science, Vol. 3, pp. 24-32.
References 569

Ford, L. R., Jr., and Fulkerson, D. R., 1956b. “Maximal Flow Through a
Network,” Can. J. of Math., Vol. 8, pp. 339-404.
Ford, L. R., Jr., and Fulkerson, D. R., 1957. “A Primal-Dual Algorithm
for the Capacitated Hitchcock Problem,” Naval Res. Logist. Quart., Vol.
4, pp. 47-54.
Ford, L. R., Jr., and Fulkerson, D. R., 1962. Flows in Networks, Princeton
Univ. Press, Princeton, N. J.
Fox, B. L., 1993. “Integrating and Accelerating Tabu Search, Simulated
Annealing, and Genetic Algorithms,” Annals of Operations Research, Vol.
41, pp. 47-67.
Fox, B. L., 1995. “Faster Simulated Annealing,” SIAM J. Optimization,
Vol. 41, pp. 47-67.
Frank, H., and Frisch, I. T., 1970. Communication, Transmission, and
Transportation Networks, Addison-Wesley, Reading, MA.
Fratta, L., Gerla, M., and Kleinrock, L., 1973. “The Flow-Deviation Method:
An Approach to Store-and-Forward Computer Communication Network
Design,” Networks, Vol. 3, pp. 97-133.
Fredman, M. L., and Tarjan, R. E., 1984. “Fibonacci Heaps and their Uses
in Improved Network Optimization Algorithms,” Proc. 25th Annual Symp.
on Found. of Comp. Sci., pp. 338-346.
Fukushima, M., 1984a. “A Modified Frank-Wolfe Algorithm for Solving
the Traffic Assignment Problem,” Transportation Research, Vol. 18B, pp.
169–177.
Fukushima, M., 1984b. “On the Dual Approach to the Traffic Assignment
Problem,” Transportation Research, Vol. 18B, pp. 235-245.
Fukushima, M., 1992. “Equivalent Differentiable Optimization Problems
and Descent Methods for Asymmetric Variational Inequalities,” Math. Pro-
gramming, Vol. 53, pp. 99-110.
Fulkerson, D. R., 1961. “An Out-of-Kilter Method for Minimal Cost Flow
Problems,” SIAM J. Appl. Math., Vol. 9, pp. 18-27.
Fulkerson, D. R., and Dantzig, G. B., 1955. “Computation of Maximum
Flow in Networks,” Naval Res. Log. Quart., Vol. 2, pp. 277-283.
Gafni, E. M., 1979. “Convergence of a Routing Algorithm,” M.S. Thesis,
Dept. of Electrical Engineering, Univ. of Illinois, Urbana, Ill.
Gafni, E. M., and Bertsekas, D. P., 1984. “Two-Metric Projection Methods
for Constrained Optimization,” SIAM J. on Control and Optimization, Vol.
22, pp. 936-964.
570 References

Gale, D., 1957. “A Theorem of Flows in Networks,” Pacific J. Math., Vol.

7, pp. 1073-1082.
Gale, D., Kuhn, H. W., and Tucker, A. W., 1951. “Linear Programming and
the Theory of Games,” in Activity Analysis of Production and Allocation,
T. C. Koopmans (ed.), Wiley, N. Y.
 
Galil, Z., 1980. “O V 5/3 E 2/3 Algorithm for the Maximum Flow Problem,”
Acta Informatica, Vol. 14, pp. 221-242.
 
Galil, Z., and Naamad, A., 1980. “O V E log2 V Algorithm for the Maxi-
mum Flow Problem,” J. of Comput. Sys. Sci., Vol. 21, pp. 203-217.
Gallager, R. G., 1977. “A Minimum Delay Routing Algorithm Using Dis-
tributed Computation,” IEEE Trans. on Communications, Vol. 23, pp. 73-
85.
Gallo, G. S., and Pallottino, S., 1982. “A New Algorithm to Find the Short-
est Paths Between All Pairs of Nodes,” Discrete Applied Mathematics, Vol.
4, pp. 23-35.
Gallo, G. S., and Pallottino, S., 1986. “Shortest Path Methods: A Unified
Approach,” Math. Programming Studies, Vol. 26, pp. 38-64.
Gallo, G. S., and Pallottino, S., 1988. “Shortest Path Algorithms,” Annals
of Operations Research, Vol. 7, pp. 3-79.
Garey, M. R., and Johnson, D. S., 1979. Computers and Intractability: A
Guide to the Theory of NP-Completeness, W. H. Freeman and Co., San
Francisco, Ca.
Gartner, N. H., 1980a. “Optimal Traffic Assignment with Elastic Demands:
A Review. Part I. Analysis Framework,” Transportation Science, Vol. 14,
pp. 174-191.
Gartner, N. H., 1980b. “Optimal Traffic Assignment with Elastic Demands:
A Review. Part II. Algorithmic Approaches,” Transportation Science, Vol.
14, pp. 192-208.
Gavish, B., Schweitzer, P., and Shlifer, E., 1977. “The Zero Pivot Phe-
nomenon in Transportation Problems and its Computational Implications,”
Math. Programming, Vol. 12, pp. 226-240.
Gelfand, S. B., and Mitter, S. K., 1989. “Simulated Annealing with Noisy
or Imprecise Measurements,” J. Opt. Theory and Applications, Vol. 69, pp.
49-62.
Geoffrion, A. M., 1970. “Elements of Large-Scale Mathematical Program-
ming, I, II,” Management Science, Vol. 16, pp. 652-675, 676-691.
Geoffrion, A. M., 1974. “Lagrangian Relaxation for Integer Programming,”
Math. Programming Studies, Vol. 2, pp. 82-114.
References 571

Gerards, A. M. H., 1995. “Matching,” Handbooks in OR and MS, Ball, M.

O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 7,
North-Holland, Amsterdam, pp. 135-224.
Gibby, D., Glover, F., Klingman, D., and Mead, M., 1983. “A Compari-
son of Pivot Selection Rules for Primal Simplex Based Network Codes,”
Operations Research Letters, Vol. 2, pp. 199-202.
Gill, P. E., Murray, W., and Wright, M. H., 1981. Practical Optimization,
Academic Press, N. Y.
Gilmore, P. C., Lawler, E. L., and Shmoys, D. B., 1985. “Well-Solved
Special Cases,” in The Traveling Salesman Problem, Lawler, E., Lenstra,
J. K., Rinnoy Kan, A. H. G., and Shmoys, D. B. (eds.), Wiley, N. Y., pp.
87-143.
Glover, F., 1986. “Future Paths for Integer Programming and Links to
Artificial Intelligence,” Computers and Operations Research, Vol. 13, pp.
533-549.
Glover, F., 1989. “Tabu Search: Part I,” ORSA J. on Computing, Vol. 1,
pp. 190-206.
Glover, F., 1990. “Tabu Search: Part II,” ORSA J. on Computing, Vol. 2,
pp. 4-32.
Glover, F., Glover, R., and Klingman, D., 1986. “The Threshold Shortest
Path Algorithm,” Math. Programming Studies, Vol. 26, pp. 12-37.
Glover, F., Glover, R., and Klingman, D., 1986. “Threshold Assignment
Algorithm,” Math. Programming Studies, Vol. 26, pp. 12-37.
Glover, F., Karney, D., and Klingman, D., 1974. “Implementation and
Computational Comparisons of Primal, Dual, and Primal-Dual Computer
Codes for Minimum Cost Network Flow Problem,” Networks, Vol. 4, pp.
191-212.
Glover, F., Karney, D., Klingman, D., and Napier, A., 1974. “A Com-
putation Study on Start Procedures, Basis Change Criteria, and Solution
Algorithms for Transportation Problems,” Management Science, Vol. 20,
pp. 793-819.
Glover, F., Klingman, D., Mote, J., and Whitman, D., 1984. “A Pri-
mal Simplex Variant for the Maximum Flow Problem,” Naval Res. Logist.
Quart., Vol. 31, pp. 41-61.
Glover, F., Klingman, D., and Phillips, N., 1985. “A New Polynomially
Bounded Shortest Path Algorithm,” Operations Research, Vol. 33, pp. 65-
73.
Glover, F., Klingman, D., and Phillips, N., 1992. Network Models in Opti-
mization and Their Applications in Practice, Wiley, N. Y.
572 References

Glover, F., Klingman, D., Phillips, N., and Schneider, R. F., 1985. “New
Polynomial Shortest Path Algorithms and Their Computational Attributes,”
Management Science, Vol. 31, pp. 1106-1128.
Glover, F., Klingman, D., and Stutz, J., 1973. “Extension of the Augmented
Predecessor Index Method to Generalized Netork Problems,” Transporta-
tion Science, Vol. 7, pp. 377-384.
Glover, F., Klingman, D., and Stutz, J., 1974. “Augmented Threaded Index
Method for Network Optimization,” INFOR, Vol. 12, pp. 293-298.
Glover, F., and Laguna, M., 1997. Tabu Search, Kluwer, Boston.
Glover, F., Taillard, E., and de Verra, D., 1993. “A User’s Guide to Tabu
Search,” Annals of Operations Research, Vol. 41, pp. 3-28.
Goffin, J. L., 1977. “On Convergence Rates of Subgradient Optimization
Methods,” Math. Programming, Vol. 13, pp. 329-347.
Goffin, J. L., Haurie, A., and Vial, J. P., 1992. “Decomposition and Non-
differentiable Optimization with the Projective Algorithm,” Management
Science, Vol. 38, pp. 284-302.
Goffin, J. L., and Kiwiel, K. C, 1996. ‘Convergence of a Simple Subgradient
Level Method,” Unpublished Report, to appear in Math. Programming.
Goffin, J. L., Luo, Z.-Q., and Ye, Y., 1993. “On the Complexity of a Column
Generation Algorithm for Convex or Quasiconvex Feasibility Problems,” in
Large Scale Optimization: State of the Art, Hager, W. W., Hearn, D. W.,
and Pardalos, P. M. (eds.), Kluwer.
Goffin, J. L., Luo, Z.-Q., and Ye, Y., 1996. “Further Complexity Analysis of
a Primal-Dual Column Generation Algorithm for Convex or Quasiconvex
Feasibility Problems,” SIAM J. on Optimization, Vol. 6, pp. 638-652.
Goffin, J. L., and Vial, J. P., 1990. “Cutting Planes and Column Generation
Techniques with the Projective Algorithm,” J. Opt. Th. and Appl., Vol. 65,
pp. 409-429.
Goldberg, A. V., 1987. “Efficient Graph Algorithms for Sequential and Par-
allel Computers,” Tech. Report TR-374, Laboratory for Computer Science,
M.I.T., Cambridge, MA.
Goldberg, A. V., 1993. “An Efficient Implementation of a Scaling Minimum-
Cost Flow Algorithm,” Proc. 3rd Integer Progr. and Combinatorial Opti-
mization Conf., pp. 251-266.
Goldberg, A. V., and Tarjan, R. E., 1986. “A New Approach to the Maxi-
mum Flow Problem,” Proc. 18th ACM STOC, pp. 136-146.
Goldberg, A. V., and Tarjan, R. E., 1990. “Solving Minimum Cost Flow
Problems by Successive Approximation,” Math. of Operations Research,
Vol. 15, pp. 430-466.
References 573

Goldberg, D. E., 1989. Genetic Algorithms in Search, Optimization, and

Machine Learning, Addison Wesley, Reading, MA.
Goldfarb, D., 1985. “Efficient Dual Simplex Algorithms for the Assignment
Problem,” Math. Programming, Vol. 33, pp. 187-203.
Goldfarb, D., and Hao, J., 1990. “A Primal Simplex Algorithm that Solves
the Maximum Flow Problem in at Most nm Pivots and O(n2 m) Time,”
Math. Programming, Vol. 47, pp. 353-365.
Goldfarb, D., Hao, J., and Kai, S., 1990a. “Anti-Stalling Pivot Rules for
the Network Simplex Algorithm,” Networks, Vol. 20, pp. 79-91.
Goldfarb, D., Hao, J., and Kai, S., 1990b. “Efficient Shortest Path Simplex
Algorithms,” Operations Research, Vol. 38, pp. 624-628.
Goldfarb, D., and Reid, J. K., 1977. “A Practicable Steepest Edge Simplex
Algorithm,” Math. Programming, Vol. 12, pp. 361-371.
Goldstein, A. A., 1967. Constructive Real Analysis, Harper and Row, N.
Y.
Gondran, M., and Minoux, M., 1984. Graphs and Algorithms, Wiley, N.
Y.
Gonzalez, R., and Rofman, E., 1985. “On Deterministic Control Problems:
An Approximation Procedure for the Optimal Cost, Parts I, II,” SIAM J.
on Control and Optimization, Vol. 23, pp. 242-285.
Graham, R. L., Lawler, E. L., Lenstra, J. K., and Rinnooy Kan, A. H. G.,
1979. “Optimization and Approximation in Deterministic Sequencing and
Scheduling: A Survey,” Annals of Discrete Math., Vol. 5, pp. 287-326.
Grötschel, M., Monma, C. L., and Stoer, M., 1995. “Design of Surviv-
able Networks,” Handbooks in OR and MS, Ball, M. O., Magnanti, T.
L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 7, North-Holland,
Amsterdam, pp. 617-672.
Grötschel, M., and Padberg, M. W., 1985. “Polyhedral Theory,” in The
Traveling Salesman Problem, Lawler, E., Lenstra, J. K., Rinnoy Kan, A.
H. G., and Shmoys, D. B. (eds.), Wiley, N. Y., pp. 251-305.
Guerriero, F., Lacagnina, V., Musmanno, R., and Pecorella, A., 1996. “Ef-
ficient Node Selection Strategies in Label-Correcting Methods for the K
Shortest Paths Problem,” Technical Report PARCOLAB No. 6/96, De-
partment of Electronics, Informatics and Systems, University of Calabria.
Guler, O., 1992. “New Proximal Point Algorithms for Convex Minimiza-
tion,” SIAM J. on Optimization, Vol. 2, pp. 649-664.
Hajek, B., 1988. “Cooling Schedules for Optimal Annealing,” Math. of
Operations Research, Vol. 13, pp. 311-329.
574 References

Hall, M., Jr., 1956. “An Algorithm for Distinct Representatives,” Amer.
Math. Monthly, Vol. 51, pp. 716-717.
Hansen, P., 1986. “The Steepest Ascent Mildest Descent Heuristic for Com-
binatorial Optimization,” Presented at the Congress on Numerical Methods
in Combinatorial Optimization, Capri, Italy.
Hearn, D. W., and Lawphongpanich, S., 1990. “A Dual Ascent Algorithm
for Traffic Assignment Problems,” Transportation Research, Vol. 24B, pp.
423-430.
Hearn, D. W., Lawphongpanish, S., and Nguyen, S., 1984. “Convex Pro-
gramming Formulation of the Asymmetric Traffic Assignment Problem,”
Transportation Research, Vol. 18B, pp. 357-365.
Hearn, D. W., Lawphongpanish, S., and Ventura, J. A., 1985. “Finiteness
in Restricted Simplicial Decomposition,” Operations Research Letters, Vol.
4, pp. 125-130.
Hearn, D. W., Lawphongpanish, S., and Ventura, J. A., 1987. “Restricted
Simplicial Decomposition: Computation and Extensions,” Math. Program-
ming Studies, Vol. 31, pp. 99-118.
Held, M., and Karp, R. M., 1970. “The Traveling Salesman Problem and
Minimum Spanning Trees,” Operations Research, Vol. 18, pp. 1138-1162.
Held, M., and Karp, R. M., 1971. “The Traveling Salesman Problem and
Minimum Spanning Trees: Part II,” Math. Programming, Vol. 1, pp. 6-25.
Helgason, R. V., and Kennington, J. L., 1977. “An Efficient Procedure for
Implementing a Dual-Simplex Network Flow Algorithm,” AIIE Transac-
tions, Vol. 9, pp. 63-68.
Helgason, R. V., and Kennington, J. L., 1995. “Primal-Simplex Algorithms
for Minimum Cost Network Flows,” Handbooks in OR and MS, Ball, M.
O., Magnanti, T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 7,
North-Holland, Amsterdam, pp. 85-133.
Helgason, R. V., Kennington, J. L., and Stewart, B. D., 1993. “The One-
to-One Shortest-Path Problem: An Empirical Analysis with the Two-Tree
Dijkstra Algorithm,” Computational Optimization and Applications, Vol.
1, pp. 47-75.
Hiriart-Urruty, J.-B., and Lemarechal, C., 1993. Convex Analysis and Min-
imization Algorithms, Vols. I and II, Springer-Verlag, Berlin and N. Y.
Hochbaum, D. S., and Shantikumar, J. G., 1990. “Convex Separable Op-
timization is not Much Harder than Linear Optimization,” J. ACM, Vol.
37, pp. 843-862.
Hoffman, A. J., 1960. “Some Recent Applications of the Theory of Lin-
ear Inequalities to Extremal Combinatorial Analysis,” Proc. Symp. Appl.
References 575

Math., Vol. 10, pp. 113-128.

Hoffman, A. J., and Kuhn, H. W., 1956. “Systems of Distinct Representa-
tives and Linear Programming,” Amer. Math. Monthly, Vol. 63, pp. 455-
460.
Hoffman, K., and Kunze, R., 1971. Linear Algebra, Prentice-Hall, Engle-
wood Cliffs, N. J.
Holloway, C. A., 1974. “An Extension of the Frank and Wolfe Method of
Feasible Directions,” Math. Programming, Vol. 6, pp. 14-27.
Hopcroft, J. E., and Karp, R. M., 1973. “A n5/2 Algorithm for Maximum
Matchings in Bipartite Graphs,” SIAM J. on Computing, Vol. 2, pp. 225-
231.
Horst, R., Pardalos, P. M., and Thoai, N. V., 1995. Introduction to Global
Optimization, Kluwer Academic Publishers, N. Y.
Hu, T. C., 1969. Integer Programming and Network Flows, Addison-Wesley,
Reading, MA.
Hung, M., 1983. “A Polynomial Simplex Method for the Assignment Prob-
lem,” Operations Research, Vol. 31, pp. 595-600.
Ibaraki, T., and Katoh, N., 1988. Resource Allocation Problems: Algorith-
mic Approaches, M.I.T. Press, Cambridge, MA.
Iri, M., 1969. Network Flows, Transportation, and Scheduling, Academic
Press, N. Y.
Iusem, A. N., Svaiter, B., and Teboulle, M., 1994. “Entropy-Like Proximal
Methods in Convex Programming,” Math. Operations Research, Vol. 19,
pp. 790-814.
Jensen, P. A., and Barnes, J. W., 1980. Network Flow Programming, Wiley,
N. Y.
Jewell, W. S., 1962. “Optimal Flow Through Networks with Gains,” Op-
erations Research, Vol. 10, pp. 476-499.
Johnson, D. B., 1977. “Efficient Algorithms for Shortest Paths in Sparse
Networks,” J. ACM, Vol. 24, pp. 1-13.
Johnson, D. S., and Papadimitriou, C. H., 1985. “Computational Com-
plexity,” in The Traveling Salesman Problem, Lawler, E., Lenstra, J. K.,
Rinnoy Kan, A. H. G., and Shmoys, D. B. (eds.), Wiley, N. Y., pp. 37-85.
Johnson, D. S., and McGeoch, L., 1997. “The Traveling Salesman Problem:
A Case Study,” in Local Search in Combinatorial Optimization, Aarts, E.,
and Lenstra, J. K. (eds.), Wiley, N. Y.
Johnson, E. L., 1966. “Networks and Basic Solutions,” Operations Re-
search, Vol. 14, pp. 619-624.
576 References

Johnson, E. L., 1972. “On Shortest Paths and Sorting,” Proc. 25th ACM
Annual Conference, pp. 510-517.
Jonker, R., and Volgenant, A., 1986. “Improving the Hungarian Assign-
ment Algorithm,” Operations Research Letters, Vol. 5, pp. 171-175.
Jonker, R., and Volgenant, A., 1987. “A Shortest Augmenting Path Al-
gorithm for Dense and Sparse Linear Assignment Problems,” Computing,
Vol. 38, pp. 325-340.
Junger, M., Reinelt, G., and Rinaldi, G., 1995. “The Traveling Sales-
man Problem,” Handbooks in OR and MS, Ball, M. O., Magnanti, T.
L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 7, North-Holland,
Amsterdam, pp. 225-330.
Karzanov, A. V., 1974. “Determining the Maximal Flow in a Network with
the Method of Preflows,” Soviet Math Dokl., Vol. 15, pp. 1277-1280.
Karzanov, A. V., and McCormick, S. T., 1997. “Polynomial Methods for
Separable Convex Optimization in Unimodular Linear Spaces with Ap-
plications to Circulations and Co-circulations in Network,” SIAM J. on
Computing, Vol. 26, pp. 1245-1275.
Kelley, J. E., 1960. “The Cutting-Plane Method for Solving Convex Pro-
grams,” J. Soc. Indust. Appl. Math., Vol. 8, pp. 703-712.
Kennington, J., and Helgason, R., 1980. Algorithms for Network Program-
ming, Wiley, N. Y.
Kennington, J., and Shalaby, M., 1977. “An Effective Subgradient Pro-
cedure for Minimal Cost Multicommodity Flow Problems,” Management
Science, Vol. 23, pp. 994-1004.
Kernighan, B. W., and Lin, S., 1970. “An Efficient Heuristic Procedure for
Partitioning Graphs,” Bell System Tech. Journal, Vol. 49, pp. 291-307.
Kershenbaum, A., 1981. “A Note on Finding Shortest Path Trees,” Net-
works, Vol. 11, pp. 399-400.
Kershenbaum, A., 1993. Network Design Algorithms, McGraw-Hill, N. Y.
Kirkpatrick, S., Gelatt, C. D., and Vecchi, M. P., 1983. “Optimization by
Simulated Annealing,” Science, Vol. 220, pp. 621-680.
Kiwiel, K. C., 1997a. “Proximal Minimization Methods with Generalized
Bregman Functions,” SIAM J. on Control and Optimization, Vol. 35, pp.
1142-1168.
Kiwiel, K. C., 1997b. “Efficiency of the Analytic Center Cutting Plane
Method for Convex Minimization,” SIAM J. on Optimization, Vol. 7, pp.
336-346.
References 577

Klee, V., and Minty, G. J., 1972. “How Good is the Simplex Algorithm?,”
in Inequalities III, O. Shisha (ed.), Academic Press, N. Y., pp. 159-175.
Klein, M., 1967. “A Primal Method for Minimal Cost Flow with Appli-
cations to the Assignment and Transportation Problems,” Management
Science, Vol. 14, pp. 205-220.
Klessig, R. W., 1974. “An Algorithm for Nonlinear Multicommodity Flow
Problems,” Networks, Vol. 4, pp. 343-355.
Klincewitz, J. C., 1989. “Implementing an Exact Newton Method for Sep-
arable Convex Transportation Problems,” Networks, Vol. 19, pp. 95-105.
König, D., 1931. “Graphok es Matrixok,” Mat. Es Fiz. Lapok, Vol. 38, pp.
116-119.
Korst, J., Aarts, E. H., and Korst, A., 1989. Simulated Annealing and
Boltzmann Machines: A Stochastic Approach to Combinatorial Optimiza-
tion and Neural Computing, Wiley, N. Y.
Kortanek, K. O., and No, H., 1993. “A Central Cutting Plane Algorithm for
Convex Semi-Infinite Programming Problems,” SIAM J. on Optimization,
Vol. 3, pp. 901-918.
Kuhn, H. W., 1955. “The Hungarian Method for the Assignment Problem,”
Naval Research Logistics Quarterly, Vol. 2, pp. 83-97.
Kumar, V., Grama, A., Gupta, A., and Karypis, G., 1994. Introduction to
Parallel Computing, Benjamin/Cummings, Redwood City, CA.
Kushner, H. J., 1990. “Numerical Methods for Continuous Control Prob-
lems in Continuous Time,” SIAM J. on Control and Optimization, Vol. 28,
pp. 999-1048.
Kushner, H. J., and Dupuis, P. G., 1992. Numerical Methods for Stochastic
Control Problems in Continuous Time, Springer-Verlag, N. Y.
Kwan Mei-Ko, 1962. “Graphic Programming Using Odd or Even Points,”
Chinese Math., Vol. 1, pp. 273-277.
Lamar, B. W., 1993. “An Improved Branch and Bound Algorithm for Min-
imum Concave Cost Network Flow Problems,” in Network Optimization
Problems, Du, D.-Z., and Pardalos, P. M. (eds.), World Scientific Publ.,
Singapore, pp. 261-287.
Land, A. H., and Doig, A. G., 1960. “An Automatic Method for Solving
Discrete Programming Problems,” Econometrica, Vol. 28, pp. 497-520.
Larsson, T., and Patricksson, M., 1992. “Simplicial Decomposition with
Disaggregated Representation for the Traffic Assignment Problem,” Trans-
portation Science, Vol. 26, pp. 4-17.
578 References

Lasdon, L. S., 1970. Optimization Theory for Large Systems, Macmillian,

N. Y.
Lawphongpanich, S., and Hearn, D., 1984. “Simplicial Decomposition of
the Asymmetric Traffic Assignment Problems,” Transportation Research,
Vol. 18B, pp. 123-133.
Lawphongpanich, S., and Hearn, D. W., 1986. “Restricted Simplicial De-
composition with Application to the Traffic Assignment Problem,” Ricerca
Operativa, Vol. 38, pp. 97-120.
Lawler, E., 1976. Combinatorial Optimization: Networks and Matroids,
Holt, Reinhart, and Winston, N. Y.
Lawler, E., Lenstra, J. K., Rinnoy Kan, A. H. G., and Shmoys, D. B., 1985.
The Traveling Salesman Problem, Wiley, N. Y.
LeBlanc, L. J., Helgason, R. V., and Boyce, D. E., 1985. “Improved Ef-
ficiency of the Frank-Wolfe Algorithm for Convex Network Programs,”
Transportation Science, Vol. 19, pp. 445–462.
LeBlanc, L. J., Morlok, E. K., and Pierskalla, W. P., 1974. “An Accurate
and Efficient Approach to Equilibrium Traffic Assignment on Congested
Networks,” Transportation Research Record, TRB-National Academy of
Sciences, Vol. 491, pp. 12-23.
LeBlanc, L. J., Morlok, E. K., and Pierskalla, W. P., 1975. “An Efficient
Approach to Solving the Road Network Equilibrium Traffic Assignment
Problem,” Transportation Research, Vol. 9, pp. 309-318.
Leventhal, T., Nemhauser, G., and Trotter, Jr., L., 1973. “A Column Gen-
eration Algorithm for Optimal Traffic Assignment,” Transportation Sci-
ence, Vol. 7, pp. 168-176.
Lemarechal, C., 1974. “An Algorithm for Minimizing Convex Functions,”
in Information Processing ’74, Rosenfeld, J. L. (ed.), North Holland Publ.
Co., Amsterdam, pp. 552-556.
Little, J. D. C., Murty, K. G., Sweeney, D. W., and Karel, C., 1963. “An
Algorithm for the Traveling Salesman Problem,” Operations Research, Vol.
11, pp. 972-989.
Lovasz, L., and Plummer, M. D., 1985. Matching Theory, North-Holland,
Amsterdam.
Luenberger, D. G., 1969. Optimization by Vector Space Methods, Wiley,
N. Y.
Luenberger, D. G., 1984. Linear and Nonlinear Programming, Addison-
Wesley, Reading, MA.
Luo, Z.-Q., 1997. “Analysis of a Cutting Plane Method that Uses Weighted
Analytic Center and Multiple Cuts,” SIAM J. of Optimization, Vol. 7, pp.
References 579

697-716.
Luo, Z.-Q., and Tseng, P., 1994. “On the Rate of Convergence of a Dis-
tributed Asynchronous Routing Algorithm,” IEEE Trans. on Automatic
Control, Vol. 39, pp. 1123-1129.
Malhotra, V. M., Kumar, M. P., and Maheshwari, S. N., 1978. “An O(|V |3 )
Algorithm for Finding Maximum Flows in Networks,” Inform. Process.
Lett., Vol. 7, pp. 277-278.
Marcotte, P., 1985. “A New Algorithm for Solving Variational Inequalities
with Application to the Traffic Assignment Problem,” Math. Programming
Studies, Vol. 33, pp. 339-351.
Marcotte, P., and Dussault, J.-P., 1987. “A Note on a Globally Convergent
Newton Method for Solving Monotone Variational Inequalities,” Opera-
tions Research Letters, Vol. 6, pp. 35-42.
Marcotte, P., and Guélat, J., 1988. “Adaptation of a Modified Newton
Method for Solving the Asymmetric Traffic Equilibrium Problem,” Trans-
portation Science, Vol. 22, pp. 112-124.
Martello, S., and Toth, P., 1990. Knapsack Problems, Wiley, N. Y.
Martinet, B., 1970. “Regularisation d’Inequations Variationnelles par Ap-
proximations Successives,” Rev. Francaise Inf. Rech. Oper., Vol. 4, pp.
154-159.
McGinnis, L. F., 1983. “Implementation and Testing of a Primal-Dual Al-
gorithm for the Assignment Problem,” Operations Research, Vol. 31, pp.
277-291.
Mendelssohn, N. S., and Dulmage, A. L., 1958. “Some Generalizations of
Distinct Representatives,” Canad. J. Math., Vol. 10, pp. 230-241.
Metropolis, N., Rosenbluth, A., Rosenbluth, M., Teller, A., and Teller, E.,
1953. “Equation of State Calculations by Fast Computing Machines,” J. of
Chemical Physisc, Vol. 21, pp. 1087-1092.
Meyer, R. R., 1979. “Two-Segment Separable Programming,” Management
Science, Vol. 25, pp. 385-395.
Miller, D., Pekny, J., and Thompson, G. L., 1990. “Solution of Large Dense
Transportation Problems Using a Parallel Primal Algorithm,” Operations
Research Letters, Vol. 9, pp. 319-324.
Minty, G. J., 1957. “A Comment on the Shortest Route Problem,” Opera-
tions Research, Vol. 5, p. 724.
Minty, G. J., 1960. “Monotone Networks,” Proc. Roy. Soc. London, A, Vol.
257, pp. 194-212.
580 References

Minieka, E., 1978. Optimization Algorithms for Networks and Graphs,

Marcel Dekker, N. Y.
Minoux, M., 1986a. Mathematical Programming: Theory and Algorithms,
Wiley, N. Y.
Minoux, M., 1986b. “Solving Integer Minimum Cost Flows with Separable
Convex Cost Objective Polynomially,” Math. Programming Studies, Vol.
26, pp. 237-239.
Minoux, M., 1989. “Network Synthesis and Optimum Network Design
Problems: Models, Solution Methods,and Applications,” Networks, Vol.
19, pp. 313-360.
Monma, C. L., and Sheng, D. D., 1986. “Backbone Network Design and
Performance Analysis: A Methodology for Packet Switching Networks,”
IEEE J. Select. Areas Comm., Vol. SAC-4, pp. 946-965.
Mulvey, J., 1978a. “Pivot Strategies for Primal-Simplex Network Codes,”
J. ACM, Vol. 25, pp. 266-270.
Mulvey, J., 1978b. “Testing a Large-Scale Network Optimization Program,”
Math. Programming, Vol. 15, pp. 291-314.
Murty, K. G., 1992. Network Programming, Prentice-Hall, Englewood Cliffs,
N. J.
Nagurney, A., 1988. “An Equilibration Scheme for the Traffic Assignment
Problem with Elastic Demands,” Transportation Research, Vol. 22B, pp.
73-79.
Nagurney, A., 1993. Network Economics: A Variational Inequality Ap-
proach, Kluwer, Dordrecht, The Netherlands.
Nemhauser, G. L., and Wolsey, L. A., 1988. Integer and Combinatorial
Optimization, Wiley, N. Y.
Nesterov, Y., 1995. “Complexity Estimates of Some Cutting Plane Methods
Based on Analytic Barrier,” Math. Programming, Vol. 69, pp. 149-176.
Nesterov, Y., and Nemirovskii, A., 1994. Interior Point Polynomial Algo-
rithms in Convex Programming, SIAM, Phila., PA.
Nguyen, S., 1974. “An Algorithm for the Traffic Assignment Problem,”
Transportation Science, Vol. 8, pp. 203-216.
Nicholson, T., 1966. “Finding the Shortest Route Between Two Points in
a Network,” The Computer Journal, Vol. 9, pp. 275-280.
Nilsson, N. J., 1971. Problem-Solving Methods in Artificial Intelligence,
McGraw-Hill, N. Y.
Nilsson, N. J., 1980. Principles of Artificial Intelligence, Tioga, Palo Alto,
CA.
References 581

O’hEigeartaigh, M., Lenstra, S. K., and Rinnoy Kan, A. H. G. (eds.), 1985.

Combinatorial Optimization: Annotated Bibliographies, Wiley, N. Y.
Ortega, J. M., and Rheinboldt, W. C., 1970. Iterative Solution of Nonlinear
Equations in Several Variables, Academic Press, N. Y.
Osman, I. H., and Laporte, G., 1996. “Metaheuristics: A Bibliography,”
Annals of Operations Research, Vol. 63, pp. 513-628.
Padberg, M. W., and Grötschel, M., 1985. “Polyhedral Computations,” in
The Traveling Salesman Problem, Lawler, E., Lenstra, J. K., Rinnoy Kan,
A. H. G., and Shmoys, D. B. (eds.), Wiley, N. Y., pp. 307-360.
Pallottino, S., 1984. “Shortest Path Methods: Complexity, Interrelations
and New Propositions,” Networks, Vol. 14, pp. 257-267.
Pallottino, S., and Scutellà, M. G., 1991. “Strongly Polynomial Algorithms
for Shortest Paths,” Ricerca Operativa, Vol. 60, pp. 33-53.
Pallottino, S., and Scutellà, M. G., 1997a. “Shortest Path Algorithms in
Transportation Models: Classical and Innovative Aspects,” Proc. of the
International Colloquium on Equilibrium in Transportation Models, Mon-
treal, Canada.
Pallottino, S., and Scutellà, M. G., 1997b. “Dual Algorithms for the Short-
est Path Tree Problem,” Networks, Vol. 29, pp. 125-133.
Pang, J.-S., 1984. “Solution of the General Multicommodity Spatial Equi-
librium Problem by Variational and Complementarity Methods,” J. of Re-
gional Science, Vol. 24, pp. 403-414.
Pang, J.-S., and Yu, C.-S., 1984. “Linearized Simplicial Decomposition
Methods for Computing Traffic Equilibria on Networks,” Networks, Vol.
14, pp. 427-438.
Papadimitriou, C. H., and Steiglitz, K., 1982. Combinatorial Optimization:
Algorithms and Complexity, Prentice-Hall, Englewood Cliffs, N. J.
Pape, U., 1974. “Implementation and Efficiency of Moore - Algorithms for
the Shortest Path Problem,” Math. Programming, Vol. 7, pp. 212-222.
Pardalos, P. M., and Rosen, J. B., 1987. Constrained Global Optimization:
Algorithms and Applications, Springer-Verlag, N. Y.
Patricksson, M., 1991. “Algorithms for Urban Traffic Network Equilibria,”
Linköping Studies in Science and Technology, Department of Mathematics,
Thesis No. 263, Linköping University, Linköping, Sweden.
Pattipati, K. R., and Alexandridis, M. G., 1990. “Application of Heuris-
tic Search and Information Theory to Sequential Fault Diagnosis,” IEEE
Trans. on Systems, Man, and Cybernetics, Vol. 20, pp. 872-887.
Pattipati, K. R., Deb, S., Bar-Shalom, Y., and Washburn, R. B., 1992. “A
582 References

New Relaxation Algorithm and Passive Sensor Data Association,” IEEE

Trans. Automatic Control, Vol. 37, pp. 198-213.
Pearl, J., 1984. Heuristics, Addison-Wesley, Reading, MA.
Peters, J., 1990. “The Network Simplex Method on a Multiprocessor,”
Networks, Vol. 20, pp. 845-859.
Phillips, C., and Zenios, S. A., 1989. “Experiences with Large Scale Net-
work Optimization on the Connection Machine,” in The Impact of Recent
Computing Advances on Operations Research, Vol. 9, Elsevier, Amster-
dam, The Netherlands, pp. 169-180.
Pinar, M. C., and Zenios, S. A., 1992. “Parallel Decomposition of Multi-
commodity Network Flows Using a Linear-Quadratic Penalty Algorithm,”
ORSA J. on Computing, Vol. 4, pp. 235-249.
Pinar, M. C., and Zenios, S. A., 1993. “Solving Nonlinear Programs with
Embedded Network Structures,” in Network Optimization Problems, Du,
D.-Z., and Pardalos, P. M. (eds.), World Scientific Publ., Singapore, pp.
177-202.
Pinar, M. C., and Zenios, S. A., 1994. “On Smoothing Exact Penalty Func-
tions for Convex Constrained Optimization,” SIAM J. on Optimization,
Vol. 4, pp. 486-511.
Pinedo, M., 1995. Scheduling: Theory, Algorithms, and Systems, Prentice-
Hall, Englewood Cliffs, N. J.
Poljak, B. T., 1987. Introduction to Optimization, Optimization Software
Inc., N. Y.
Polymenakos, L. C., 1995. “
-Relaxation and Auction Algorithms for the
Convex Cost Network Flow Problem,” Ph.D. Thesis, Electrical Engineering
and Computer Science Dept, M.I.T., Cambridge, MA.
Polymenakos, L. C., and Bertsekas, D. P., 1994. “Parallel Shortest Path
Auction Algorithms,” Parallel Computing, Vol. 20, pp. 1221-1247.
Polymenakos, L. C., Bertsekas, D. P., and Tsitsiklis, J. N., 1998. “Efficient
Algorithms for Continuous-Space Shortest Path Problems,” IEEE Trans.
on Automatic Control, Vol. AC-43, pp. 278-283.
Poore, A. B., 1994. “Multidimensional Assignment Formulation of Data
Association Problems Arising from Multitarget Tracking and Multisensor
Data Fusion,” Computational Optimization and Applications, Vol. 3, pp.
27-57.
Poore, A. B., and Robertson, A. J. A., 1997. New Lagrangian Relaxation
Based Algorithm for a Class of Multidimensional Assignment Problems,”
Computational Optimization and Applications, Vol. 8, pp. 129-150.
Powell, W. B., Jaillet, P., and Odoni, A., 1995. “Stochastic and Dynamic
References 583

Networks and Routing,” Handbooks in OR and MS, Ball, M. O., Magnanti,

T. L., Monma, C. L., and Nemhauser, G. L. (eds.), Vol. 8, North-Holland,
Amsterdam, pp. 141-295.
Powell, W. B., Berkkam, E., and Lustig, I. J., 1993. “On Algorithms for
Nonlinear Dynamic Networks,” in Network Optimization Problems, Du,
D.-Z., and Pardalos, P. M. (eds.), World Scientific Publ., Singapore, pp.
177-202.
Pulleyblank, W., 1983. “Polyhedral Combinatorics,” in Mathematical Pro-
gramming: The State of the Art - Bonn 1982, by Bachem, A., Grötschel,
M., and Korte, B., (eds.), Springer, Berlin, pp. 312-345.
Pulleyblank, W., Cook, W., Cunningham, W., and Schrijver, A., 1993. An
Introduction to Combinatorial Optimization, Wiley, N. Y.
Resende, M. G. C., and Veiga, G., 1993. “An Implementation of the Dual
Affine Scaling Algorithm for Minimum-Cost Flow on Bipartite Uncapaci-
tated Networks,” SIAM J. on Optimization, Vol. 3, pp. 516-537.
Resende, M. G. C., and Pardalos, P. M., 1996. “Interior Point Algorithms
for Network Flow Problems,” Advances in Linear and Integer Program-
ming, Oxford Lecture Ser. Math. Appl., Vol. 4, Oxford Univ. Press, New
York, pp. 145-185.
Rockafellar, R. T., 1967. “Convex Programming and Systems of Elemen-
tary Monotonic Relations,” J. of Math. Analysis and Applications, Vol. 19,
pp. 543-564.
Rockafellar, R. T., 1969. “The Elementary Vectors of a Subspace of RN ,”
in Combinatorial Mathematics and its Applications, by Bose, R. C., and
Dowling, T. A. (eds.), University of North Carolina Press, pp. 104-127.
Rockafellar, R. T., 1970. Convex Analysis, Princeton Univ. Press, Prince-
ton, N. J.
Rockafellar, R. T., 1976. “Monotone Operators and the Proximal Point
Algorithm,” SIAM J. on Control and Optimization, Vol. 14, pp. 877-898.
Rockafellar, R. T., 1981. “Monotropic Programming: Descent Algorithms
and Duality,” in Nonlinear Programming 4, by Mangasarian, O. L., Meyer,
R. R., and Robinson, S. M. (eds.), Academic Press, N. Y., pp. 327-366.
Rockafellar, R. T., 1984. Network Flows and Monotropic Programming,
Wiley, N. Y.
Rudin, W., 1976. Real Analysis, McGraw Hill, N. Y.
Sahni, S., and Gonzalez, T., 1976. “P -Complete Approximation Problems,”
J. ACM, Vol. 23, pp. 555-565.
Schwartz, B. L., 1994. “A Computational Analysis of the Auction Algo-
rithm,” Eur. J. of Operations Research, Vol. 74, pp. 161-169.
584 References

Sheffi, Y., 1985. Urban Transportation Networks. Equilibrium Analy-

sis with Mathematical Programming Methods, Prentice-Hall, Englewood
Cliffs, N. J.
Shier, D. R., 1979. “On Algorithms for Finding the K Shortest Paths in a
Network,” Networks, Vol. 9, pp. 195-214.
Shier, D. R., and Witzgall, C., 1981. “Properties of Labeling Methods for
Determining Shortest Path Trees,” J. Res. Natl. Bureau of Standards, Vol.
86, pp. 317-330.
Shiloach, Y., and Vishkin, U., 1982. “An O(n2 log n) Parallel Max-Flow
Algorithm,” J. Algorithms, Vol. 3, pp. 128-146.
Schrijver, A., 1986. Theory of Linear and Integer Programming, Wiley, N.
Y.
Shapiro, J. E., 1979. Mathematical Programming Structures and Algo-
rithms, Wiley, N. Y.
Shor, N. Z., 1985. Minimization Methods for Nondifferentiable Functions,
Springer-Verlag, Berlin.
Srinivasan, V., and Thompson, G. L., 1973. “Benefit-Cost Analysis of Cod-
ing Techniques for Primal Transportation Algorithm,” J. ACM, Vol. 20, pp.
194-213.
Strang, G., 1976. Linear Algebra and Its Applications, Academic Press, N.
Y.
Suchet, C., 1949. Electrical Engineering, Vol. 68, pp. 843-844.
Tabourier, Y., 1973. “All Shortest Distances in a Graph: An Improvement
to Dantzig’s Inductive Algorithm,” Disc. Math., Vol. 4, pp. 83-87.
Tardos, E., 1985. “A Strongly Polynomial Minimum Cost Circulation Al-
gorithm,” Combinatorica, Vol. 5, pp. 247-255.
Teboulle, M., 1992. “Entropic Proximal Mappings with Applications to
Nonlinear Programming,” Math. of Operations Research, Vol. 17, pp. 1-
21.
Toint, P. L., and Tuyttens, D., 1990. “On Large Scale Nonlinear Network
Optimization,” Math. Programming, Vol. 48, pp. 125-159.
Tseng, P., 1986. “Relaxation Methods for Monotropic Programming Prob-
lems,” Ph.D. Thesis, Dept. of Electrical Engineering and Computer Science,
M.I.T., Cambridge, MA.
Tseng, P., 1991. “Relaxation Method for Large Scale Linear Programming
Using Decomposition,” Math. of Operations Research, Vol. 17, pp. 859-880.
Tseng, P., 1998. “An
-Out-of-Kilter Method for Monotropic Program-
ming,” Department of Mathematics Report, Univ. of Washington, Seattle,
References 585

Wash.
Tseng, P., and Bertsekas, D. P., 1987. “Relaxation Methods for Linear
Programs,” Math. of Operations Research, Vol. 12, pp. 569-596.
Tseng, P., and Bertsekas, D. P., 1990. “Relaxation Methods for Monotropic
Programs,” Math. Programming, Vol. 46, 1990, pp. 127-151.
Tseng, P., and Bertsekas, D. P., 1993. “On the Convergence of the Expo-
nential Multiplier Method for Convex Programming,” Math. Programming,
Vol. 60, pp. 1-19.
Tseng, P., and Bertsekas, D. P., 1996. “An Epsilon-Relaxation Method
for Separable Convex Cost Generalized Network Flow Problems,” Lab. for
Information and Decision Systems Report P-2374, M.I.T., Cambridge, MA.
Tseng, P., Bertsekas, D. P., and Tsitsiklis, J. N., 1990. “Partially Asyn-
chronous Parallel Algorithms for Network Flow and Other Problems,”
SIAM J. on Control and Optimization, Vol. 28, pp. 678-710.
Tsitsiklis, J. N., 1989. “Markov Chains with Rare Transitions and Simu-
lated Annealing,” Math. of Operations Research, Vol. 14, pp. 70-90.
Tsitsiklis, J. N., 1992. “Special Cases of Traveling Salesman and Repairman
Problems with Time Windows,” Networks, Vol. 22, pp. 263-282.
Tsitsiklis, J. N., 1995. “Efficient Algorithms for Globally Optimal Trajec-
tories,” IEEE Trans. on Automatic Control, Vol. 40, pp. 1528-1538.
Tsitsiklis, J. N., and Bertsekas, D. P., 1986. “Distributed Asynchronous
Optimal Routing in Data Networks,” IEEE Trans. on Automatic Control,
Vol. 31, pp. 325-331.
Ventura, J. A., and Hearn, D. W., 1993. “Restricted Simplicial Decomposi-
tion for Convex Constrained Problems,” Math. Programming, Vol. 59, pp.
71-85.
Voß, S., 1992. “Steiner’s Problem in Graphs: Heuristic Methods,”, Discrete
Applied Math., Vol. 40, pp. 45-72.
Von Randow, R., 1982. Integer Programming and Related Areas: A Clas-
sified Bibliography 1978-1981, Lecture Notes in Economics and Mathemat-
ical Systems, Vol. 197, Springer-Verlag, N. Y.
Von Randow, R., 1985. Integer Programming and Related Areas: A Clas-
sified Bibliography 1982-1984, Lecture Notes in Economics and Mathemat-
ical Systems, Vol. 243, Springer-Verlag, N. Y.
Warshall, S., 1962. “A Theorem on Boolean Matrices,” J. ACM, Vol. 9,
pp. 11-12.
Wein, J., and Zenios, S. A., 1991. “On the Massively Parallel Solution of
the Assignment Problem,” J. of Parallel and Distributed Computing, Vol.
586 References

13, pp. 228-236.

Whitting, P. D., and Hillier, J. A., 1960. “A Method for Finding the Short-
est Route Through a Road Network,” Operations Research Quart., Vol.
11, pp. 37-40.
Winter, P., 1987. “Steiner Problem in Networks: A Survey,” Networks, Vol.
17, pp. 129-167.
Wright, S. J., 1997. Primal-Dual Interior Point Methods, SIAM, Phila.,
PA.
Ye, Y., 1992. “A Potential Reduction Algorithm Allowing Column Gener-
ation,” SIAM J. on Optimization, Vol. 2, pp. 7-20.
Ye, Y., 1997. Interior Point Algorithms: Theory and Analysis, Wiley, N.
Y.
Zadeh, N., 1973a. “A Bad Network Problem for the Simplex Method and
Other Minimum Cost Flow Algorithms,” Math. Programming, Vol. 5, pp.
255-266.
Zadeh, N., 1973b. “More Pathological Examples for Network Flow Prob-
lems,” Math. Programming, Vol. 5, pp. 217-224.
Zadeh, N., 1979. “Near Equivalence of Network Flow Algorithms,” Tech-
nical Report No. 26, Dept. of Operations Research, Stanford University,
CA.
Zenios, S. A., and Mulvey, J. M., 1986. “Relaxation Techniques for Strictly
Convex Network Problems,” Annals of Operations Research, Vol. 5, pp.
517-538.
Zoutendijk, G., 1976. Mathematical Programming Methods, North Hol-
land, Amsterdam.