0% found this document useful (0 votes)
29 views

Network Optimization: Networks

This document summarizes key concepts in network optimization including: 1) Network flows are represented by flow vectors where each element corresponds to flow on an arc. The flows must satisfy balance equations at each node given by an incidence matrix and external supply vector. 2) Minimum cost network flow problems aim to route flow through a network at minimum cost while respecting capacity constraints on each arc. 3) Maximum flow problems aim to route the maximum amount of flow between a source and sink node in a network. This can be formulated as a special case of minimum cost flow. 4) Shortest path problems find the minimum length path between nodes, where extreme optimal solutions correspond to single unit flows on arcs of the

Uploaded by

Gsadeghy
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
29 views

Network Optimization: Networks

This document summarizes key concepts in network optimization including: 1) Network flows are represented by flow vectors where each element corresponds to flow on an arc. The flows must satisfy balance equations at each node given by an incidence matrix and external supply vector. 2) Minimum cost network flow problems aim to route flow through a network at minimum cost while respecting capacity constraints on each arc. 3) Maximum flow problems aim to route the maximum amount of flow between a source and sink node in a network. This can be formulated as a special case of minimum cost flow. 4) Shortest path problems find the minimum length path between nodes, where extreme optimal solutions correspond to single unit flows on arcs of the

Uploaded by

Gsadeghy
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

EE236A (Fall 2007-08)

Lecture 8
Network optimization
network ows
minimum cost network ow problem
extreme ows
shortest path problem
bipartite matching
81
Networks
network (directed graph): m nodes connected by n directed arcs
arcs are ordered pairs (i, j)
we assume there is at most one arc from node i to node j
we assume there are no self-loops (arcs (i, i))
arc-node incidence matrix A R
mn
:
A
ij
=
_
_
_
1 arc j starts at node i
1 arc j ends at node i
0 otherwise
column sums of A are zero: 1
T
A = 0
reduced arc-node incidence matrix

A R
(m1)n
: the matrix formed
by the rst m1 rows of A
Network optimization 82
example (m = 6, n = 8)
1
2
3
4 5
6
1
2
3
4
5
6
7
8
A =
_

_
1 1 0 0 0 0 0 1
1 0 1 0 0 0 0 1
0 1 1 1 1 1 0 0
0 0 0 1 0 0 1 0
0 0 0 0 0 1 1 0
0 0 0 0 1 0 0 0
_

_
Network optimization 83
Network ow
ow vector x R
n
x
j
: ow (of material, trac, charge, information, . . . ) through arc j
positive if in direction of arc; negative otherwise
total ow leaving node i:
n

j=1
A
ij
x
j
= (Ax)
i
i
x
j
A
ij
= 1
x
k
A
ik
= 1
Network optimization 84
External supply
supply vector b R
m
b
i
: external supply at node i
negative b
i
represents external demand from the network
must satisfy 1
T
b = 0 (total supply = total demand)
i
x
j
A
ij
= 1
x
k
A
ik
= 1
b
i
balance equations: Ax = b
reduced balance equations:

Ax = (b
1
, . . . , b
m1
)
Network optimization 85
Minimum cost network ow problem
minimize c
T
x
subject to Ax = b
l x u
c
i
is unit cost of ow through arc i
l
j
and u
j
are limits on ow through arc j (typically, l
j
0, u
j
0)
we assume l
j
< u
j
, but allow l
j
= and u
j
=
includes many network problems as special cases
Network optimization 86
Max ow problem
maximize ow between node 1 (source) and node m (sink)
1
m
t t
maximize t
subject to Ax = te
l x u
where e = (1, 0, . . . , 0, 1)
Network optimization 87
interpretation as minimum cost ow problem
1
m
articial arc n + 1
minimize t
subject to
_
A e

_
x
t
_
= 0
l x u
Network optimization 88
Paths and cycles
path from node s to node t: sequence of arcs P
1
, . . . , P
N
P
k
= (i
k1
, i
k
) or P
k
= (i
k
, i
k1
), i
0
= s, i
N
= t
example (page 83): arcs 1, 3, 4, 7 form a path from node 1 to node 5
directed path sequence of arcs P
1
, . . . , P
N
P
k
= (i
k1
, i
k
) i
0
= s, i
N
= t
example: arcs 1, 3, 6 form a directed path from node 1 to node 5
(directed) cycle: (directed) path from a node to itself
example: arcs 1, 2, 3 form a cycle; arcs 4, 6, 7 form a directed cycle
Network optimization 89
Acyclic networks and trees
connected network: there exists a path between every pair of nodes
acyclic network: does not contain cycles
tree: connected acyclic network
connected, not acyclic
acyclic, not connected
tree
Network optimization 810
Topology and rank of incidence matrix
network is connected if and only if
rankA = rank

A = m1
Ax = b is solvable for all b with 1
T
b = 0
network is acyclic if and only if
rankA = rank

A = n
if Ax = b is solvable, its solution is unique
network is a tree if and only if
rank(A) = rank

A = n = m1
Ax = b has a unique solution for all b with 1
T
b = 0
Network optimization 811
Solving balance equations for tree networks
1
2
3
4
5
1
2 3
4
_

_
1 0 0 0
0 1 0 0
1 1 1 0
0 0 0 1
0 0 1 1
_

_
_

_
x
1
x
2
x
3
x
4
_

_
=
_

_
b
1
b
2
b
3
b
4
b
5
_

_
x
1
= b
1
, x
2
= b
2
, x
3
= b
1
b
2
b
3
, x
4
= b
4
in general, choose node m as root node and take
x
j
=

nodes i downstream of arc j


b
i
important consequence: x Z
n
if b Z
m
Network optimization 812
Solving balance equations for acyclic networks
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
x
1
= b
1
x
2
= b
2
x
3
= b
3
x
4
= b
4
x
5
= b
5
x
6
= b
1
b
2
b
6
x
7
= b
7
x
8
= b
3
b
4
b
5
b
8
can solve using only additions/subtractions
x Z
n
if b Z
m
Network optimization 813
Integrality of extreme ows
P is polyhedron of feasible ows
Ax = b, l x u
we will show that the extreme points of P are integer vectors if
the external supplies b
i
are integer
the ow limits l
i
, u
i
are integer (or )
Network optimization 814
proof. suppose x is an extreme ow with
l
j
< x
j
< u
j
, j = 1, . . . , K, x
j
=
_
l
j
j = K + 1, . . . , L
u
j
j = L + 1, . . . , n
we prove that x
1
,. . . , x
K
are integers
1. apply rank test of page 319 to the inequalities
l x u, Ax b, Ax b
rank test:
rank
_
_
_
_
_

_
0 I 0
0 0 I
B
0
B

B
+
B
0
B

B
+
_

_
_
_
_
_
= n
where A =
_
B
0
B

B
+

, B
0
R
mK
, etc.
conclusion: rankB
0
= K (subnetwork with arcs 1, . . . , K is acyclic)
Network optimization 815
2. y = (x
1
, . . . , x
K
) satises
B
0
y = b
_
B

B
+

_
_
x
K+1
.
.
.
x
n
_
_
(1)
interpretation: balance equations of an acyclic subnetwork with
incidence matrix B
0
, ow vector y, and integer external supplies
b
_
B

B
+

_
_
x
K+1
.
.
.
x
n
_
_
conclusion (from page 813): y is an integer vector
Network optimization 816
example (l
i
= 0, u
i
= for all arcs)
1
2
3
4 5
6
b
1
= 2
b
2
= 1
b
3
= 3
b
4
= 2 b
5
= 2
b
6
= 0
1
2
3
4
5
6
7
8
x = (0, 2, 1, 0, 0, 0, 2, 0) is an extreme ow:
it is feasible
subgraph with arcs 2, 3, 7 is acyclic
Network optimization 817
Shortest path problem
minimize 1
T
x
subject to Ax = (1, 0, . . . , 0, 1)
0 x 1
extreme optimal solutions satisfy x
i
{0, 1}
arcs with x
i
= 1 form a shortest (forward) path between nodes 1 and m
extends to networks with non-unit positive lengths
can be solved very eciently via specialized algorithms
Network optimization 818
Weighted (bipartite) matching
match N people to N tasks
each person assigned to one task; each task assigned to one person
cost of matching person i to task j is a
ij
minimum cost ow formulation
min.

N
i,j=1
a
ij
x
ij
s.t.

N
i=1
x
ij
= 1, j = 1, . . . , N

N
j=1
x
ij
= 1, i = 1, . . . , N
0 x
ij
1, i, j = 1, . . . , N
example (N = 3)
1
2
3
1
2
3
1
1
1
1
1
1
integrality: extreme optimal solution satises x
ij
{0, 1}
Network optimization 819

You might also like