Exact and Heuristic Methods in Combinatorial Optimization 2nd Edition Rafael Martí Instant Download
Exact and Heuristic Methods in Combinatorial Optimization 2nd Edition Rafael Martí Instant Download
https://ebookmeta.com/product/exact-and-heuristic-methods-in-
combinatorial-optimization-2nd-edition-rafael-marti/
https://ebookmeta.com/product/multi-objective-combinatorial-
optimization-problems-and-solution-methods-1st-edition-mehdi-
toloo-editor/
https://ebookmeta.com/product/first-order-methods-in-
optimization-1st-edition-amir-beck/
https://ebookmeta.com/product/matrix-numerical-and-optimization-
methods-in-science-and-engineering-1st-edition-kevin-w-cassel/
https://ebookmeta.com/product/tribe-space-and-mobilisation-
colonial-dynamics-and-post-colonial-dilemma-in-tribal-
studies-1st-edition-maguni-charan-behera/
Bullet Savage Crows MC Texas Charter 8 1st Edition T O
Smith
https://ebookmeta.com/product/bullet-savage-crows-mc-texas-
charter-8-1st-edition-t-o-smith/
https://ebookmeta.com/product/writings-on-the-wall-arlissa-s-
story-goddess-a-brouette/
https://ebookmeta.com/product/just-a-little-prick-2nd-edition-
hilary-butler-peter-butler/
https://ebookmeta.com/product/a-place-called-home-1st-edition-
david-ambroz/
https://ebookmeta.com/product/colton-the-boys-of-brooke-hollow-
academy-2-1st-edition-charlie-jules/
Campbell's Operative Orthopaedics 4-Volume Set 14th
Edition Frederick M. Azar
https://ebookmeta.com/product/campbells-operative-
orthopaedics-4-volume-set-14th-edition-frederick-m-azar/
Applied Mathematical Sciences
Rafael Martí
Gerhard Reinelt
Exact and
Heuristic Methods
in Combinatorial
Optimization
A Study on the Linear Ordering and
the Maximum Diversity Problem
Second Edition
Applied Mathematical Sciences
Volume 175
Series Editors
Anthony Bloch, Department of Mathematics, University of Michigan, Ann Arbor,
MI, USA
C. L. Epstein, Department of Mathematics, University of Pennsylvania,
Philadelphia, PA, USA
Alain Goriely, Department of Mathematics, University of Oxford, Oxford, UK
Leslie Greengard, New York University, New York, NY, USA
Advisory Editors
J. Bell, Center for Computational Sciences and Engineering, Lawrence Berkeley
National Laboratory, Berkeley, CA, USA
P. Constantin, Department of Mathematics, Princeton University, Princeton, NJ,
USA
R. Durrett, Department of Mathematics, Duke University, Durham, CA, USA
R. Kohn, Courant Institute of Mathematical Sciences, New York University,
New York, NY, USA
R. Pego, Department of Mathematical Sciences, Carnegie Mellon University,
Pittsburgh, PA, USA
L. Ryzhik, Department of Mathematics, Stanford University, Stanford, CA, USA
A. Singer, Department of Mathematics, Princeton University, Princeton, NJ, USA
A. Stevens, Department of Applied Mathematics, University of Münster, Münster,
Germany
S. Wright, Computer Sciences Department, University of Wisconsin, Madison, WI,
USA
Founding Editor
F. John, New York University, New York, NY, USA
J. P. LaSalle, Brown University, Providence, RI, USA
L. Sirovich, Brown University, Providence, RI, USA
The mathematization of all sciences, the fading of traditional scientific boundaries,
the impact of computer technology, the growing importance of computer modeling
and the necessity of scientific planning all create the need both in education and
research for books that are introductory to and abreast of these developments. The
purpose of this series is to provide such books, suitable for the user of mathematics,
the mathematician interested in applications, and the student scientist. In particular,
this series will provide an outlet for topics of immediate interest because of the
novelty of its treatment of an application or of mathematics being applied or lying
close to applications. These books should be accessible to readers versed in
mathematics or science and engineering, and will feature a lively tutorial style, a
focus on topics of current interest, and present clear exposition of broad appeal.
A compliment to the Applied Mathematical Sciences series is the Texts in Applied
Mathematics series, which publishes textbooks suitable for advanced undergraduate
and beginning graduate courses.
123
Rafael Martí Gerhard Reinelt
Departamento de Estadística e Department of Computer Science
Investigación Operativa University of Heidelberg
Universitat de València Heidelberg, Germany
Valencia, Spain
This Springer imprint is published by the registered company Springer-Verlag GmbH, DE part of
Springer Nature.
The registered company address is: Heidelberger Platz 3, 14197 Berlin, Germany
To Mariola for her life, and to Mila for giving
life to her.
Rafa Martı́
Preface
Faced with the challenge of solving hard optimization problems that abound in the
real world, classical methods often encounter serious difficulties. Important applica-
tions in business, engineering or economics cannot be tackled by the solution meth-
ods that have been the predominant focus of academic research throughout the past
four decades. Exact and heuristic approaches are dramatically changing our ability
to solve problems of practical significance and are extending the frontier of prob-
lems that can be handled effectively. In this text we describe state-of-the-art opti-
mization methods, both exact and heuristic, for two difficult optimization problems:
the linear ordering problem (LOP), and the maximum diversity problem (MDP).
In this way, we provide the reader with the background, elements and strategies to
tackle a wide range of different combinatorial optimization problems.
The idea for writing the first edition of this book came up when the authors
met at the University of Valencia in 2005. While comparing our experiences with
regard to various aspects of the LOP, we realized that most of the optimization
technologies had been successfully applied to solve this problem. We also found that
there were only a small number of books covering all state-of-the-art optimization
methods for hard optimization problems (especially considering both exact methods
and heuristics together). We thought that the LOP would make an ideal example to
survey these methods applied to one problem and felt the time was ripe to embark
on the project of writing a monograph.
In this second edition, we performed a major revision of the book, to offer the
reader a wider approach to combinatorial optimization. In particular, we included a
second problem, the maximum diversity problem (MDP), to describe the optimiza-
tion methods covered in the book both in the LOP and in the MDP. Considering that
in the LOP solutions are usually represented with permutations, and in the MDP
with binary variables, the adaptation of the different optimization technologies to
these two problems, provide readers with more elements, tools, and search strate-
gies to create their own solving methods
This textbook is devoted to the LOP and the MDP, their origins, applications,
instances and especially to methods for their effective approximate or exact solution.
Our intention is to provide basic principles and fundamental ideas and reflect the
vii
viii Preface
state-of-the-art of heuristic and exact methods, thus allowing the reader to create his
or her personal successful applications of the solution methods. The book is meant
to be of interest for researchers and practitioners in computer science, mathematics,
operations research, management science, industrial engineering, and economics. It
can be used as a textbook on issues of practical optimization in a master’s course or
as a reference resource for engineering optimization algorithms.
To make the book accessible to a wider audience, it is to a large extent self-
contained, providing the reader with the basic definitions and concepts in optimiza-
tion. However, in order to limit the size of this monograph we have not included
extensive introductions. Readers interested in further details are referred to appro-
priate textbooks such as [4, 103, 129, 155, 156, 163].
The structure of this book is as follows. Chapter 1 provides an introduction to the
two problems and their applications, and describes the sets of benchmark instances
which we are using for our computational experiments and which have been made
publically available. Chapter 2 describes basic heuristic methods such as construc-
tion and local searches. Chapter 3 expands on Chapter 2 and covers meta-heuristics
in which the simple methods are now embedded in complex solution algorithms
based on different paradigms, such as evolution or learning strategies. Chapter 4
discusses branch-and-bound, the principal approach for solving difficult problems
to optimality. A special version based on polyhedral combinatorics, branch-and-cut,
is presented in Chapter 5. Chapter 6 deals in more detail with the linear ordering
polytope which is at the core of branch-and-cut algorithms. The book concludes
with Chapter 7, where a number of further aspects of the LOP and MDP, and poten-
tial issues for further research are described.
Rafael Martı́’s research was partially supported by the Ministerio de Ciencia e
Innovación of Spain (Grant Ref. PGC2018-0953322-B-C21 /MCIU/AEI/FEDER-
UE).
We are in debt to many people, but in particular to some very good friends and
colleagues who helped us to gain a deeper understanding of our two problems: Vi-
cente Campos, Thomas Christof, Angel Corberán, Abraham Duarte, Fred Glover,
Martin Grötschel, Michael Jünger, Manuel Laguna, Anna Martı́nez-Gavara, Fran-
cisco Parreño, and Mauricio Resende.
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The Linear Ordering Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 The LOLIB Library of Benchmark Instances . . . . . . . . . . . . . 10
1.2 The Maximum Diversity Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.1 Diversity measures and models . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.2.2 The MDPLIB Library of Benchmark Instances . . . . . . . . . . . 22
2 Heuristic Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1.1 Assessing the Quality of Heuristics . . . . . . . . . . . . . . . . . . . . . 29
2.2 Construction Heuristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.1 Early LOP Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.2 Reconstruction in the MDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.3 General Insertions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2.4 Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3 Local Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.1 Insertions in permutation problems . . . . . . . . . . . . . . . . . . . . . 39
2.3.2 Exchanges in binary problems . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.3 LOP methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4 Multi-Start Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.1 Variants of Multi-Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.2 Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3 Meta-Heuristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 GRASP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2.1 Construction Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.2 Improvement Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.3 Strategic Oscillation and Iterated Greedy . . . . . . . . . . . . . . . . . . . . . . . 73
3.4 Tabu Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
ix
x Contents
4 Branch-and-Bound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.2 Branch-and-Bound with Partial Orderings for the LOP . . . . . . . . . . . 127
4.3 Lexicographic Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.4 Extension of Lexicographic Search to Branch-and-Bound . . . . . . . . . 129
4.5 Branch-and-Bound with Lagrangean Relaxation . . . . . . . . . . . . . . . . . 131
4.6 Improved MDP formulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.7 Branch-and-Bound with Partial Selections for the MDP . . . . . . . . . . 138
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Chapter 1
Introduction
Abstract This chapter introduces the basic definitions, main elements, applications,
and instances of two optimization problems, the linear ordering problem (LOP), and
the maximum diversity problem (MDP). We will use them in the next chapters to
describe heuristics, meta-heuristics and exact approaches, and to report our experi-
ments.
The LOP is one of the classical combinatorial optimization problems which
was already classified as NP-hard in 1979 by Garey and Johnson [64]. It has re-
ceived considerable attention in various application areas ranging from archeology
and scheduling to economics. Solution methods for the LOP have been proposed
since 1958, when Chenery and Watanabe outlined some ideas on how to obtain so-
lutions for this problem. The interest in this problem has continued over the years,
resulting in the book [143] and many recent papers in scientific journals. This chap-
ter surveys the main LOP applications and benchmark library of instances LOLIB.
The challenge of maximizing the diversity of a collection of points arises in
a variety of settings, from location to genetics. The growing interest of dealing
with diversity translated into mathematical models and computer algorithms in
the late eighties, when Michael Kuby studied dispersion maximization in general
graphs [54]. The MDP is the first model proposed and the most studied to deal with
diversity, and was classified as NP-hard. Many optimization methods have been
proposed to obtain efficient solutions to this problem, which makes it especially
convenient as an illustrative example in this book. This chapter surveys the differ-
ent models proposed to maximize diversity, their applications, and the benchmark
library of instances MDPLIB.
In its graph version the LOP is defined as follows. Let Dn = (Vn , An ) denote the com-
plete digraph on n nodes, i.e., the directed graph with node set Vn = {1, 2, . . . , n} and
the property that for every pair of nodes i and j there is an arc (i, j) from i to j and
Alternatively, the LOP can be defined as a matrix problem, the so-called trian-
gulation problem.
Triangulation problem
Let an (n, n)-matrix H = (Hi j ) be given. Determine a simultaneous per-
mutation of the rows and columns of H such that the sum of superdiag-
onal entries becomes as large as possible.
Obviously, by setting arc weights ci j = Hi j for the complete digraph Dn , the trian-
gulation problem for H can be solved as a linear ordering problem in Dn . Conversely,
a linear odering problem for Dn can be transformed to a triangulation problem for
an (n, n)-matrix H by setting Hi j = ci j and the diagonal entries Hii = 0.
Consider as an example the (5,5)-matrix
1.1 The Linear Ordering Problem 3
0 16 11 15 7
21 0 14 15 9
H =
26 23 0 26 12.
22 22 11 0 13
30 28 25 24 0
1.1.1 Applications
The acyclic subdigraph problem (ASP) is defined as follows. Given a directed graph
D = (V, A) with arc weights di j , for all (i, j) ∈ A, determine a subset B ⊆ A which
contains no directed cycles and has maximum weight d(B) = ∑(i, j)∈B di j .
It can easily be seen that this problem is equivalent to the LOP. For a given ASP
define a LOP on Dn , where n = |V |, by setting for 1 ≤ i, j ≤ n, i 6= j:
(
max{0, di j }, if (i, j) ∈ A,
ci j =
0, otherwise.
Fig. 1.1 shows a digraph on 9 nodes where the arcs of a minimum feedback arc
set are drawn as dotted lines. If the six arcs of the feedback arc set are removed, we
obtain an acyclic arc set.
There are some further problems dealing with acyclic subdigraphs. The node in-
duced acyclic subdigraph problem asks for a node set W ⊆ V such that the subdi-
graph (W, A(W )) is acyclic. (Here A(W ) denotes the set of arcs with both end nodes
in W .) The problem can be defined either with node weights d, and d(W ) is to be
maximized, or with arc weights c where c(A(W )) has to be maximum. Analogously,
the feedback node set problem is to find a set W ⊆ V such that (V \W, A(V \W )) is
acyclic. Here, sums of node weights or arc weights have to be minimized.
The request that solution digraphs have to be node induced adds a further com-
plexity. These problems cannot be transformed to a pure linear ordering problem
and are even more difficult.
Linear ordering problems may occur whenever rankings of some objects are to
be determined. Consider for example the following situation. A set of n objects
O1 , O2 , . . . , On is given which have to be rated by m persons according to their in-
dividual preferences. Then a ranking of these objects is to be found which reflects
these single rankings as closely as possible. The first question to be answered is
how the individual rankings can be obtained. One solution is a pairwise comparison
experiment. For any pair Oi and O j , 1 ≤ i < j ≤ n, of objects each person decides
whether Oi should be preferred to O j or vice versa. The results of these m n2 com-