doc/min_cost_flow.dox
author Peter Kovacs <kpeter@inf.elte.hu>
Tue, 15 Mar 2011 19:32:21 +0100
changeset 1047 ddd3c0d3d9bf
parent 835 c92296660262
child 1164 f63ba40a60f4
permissions -rw-r--r--
Implement the scaling Price Refinement heuristic in CostScaling (#417)
instead of Early Termination.

These two heuristics are similar, but the newer one is faster
and not only makes it possible to skip some epsilon phases, but
it can improve the performance of the other phases, as well.
kpeter@710
     1
/* -*- mode: C++; indent-tabs-mode: nil; -*-
kpeter@710
     2
 *
kpeter@710
     3
 * This file is a part of LEMON, a generic C++ optimization library.
kpeter@710
     4
 *
alpar@956
     5
 * Copyright (C) 2003-2010
kpeter@710
     6
 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
kpeter@710
     7
 * (Egervary Research Group on Combinatorial Optimization, EGRES).
kpeter@710
     8
 *
kpeter@710
     9
 * Permission to use, modify and distribute this software is granted
kpeter@710
    10
 * provided that this copyright notice appears in all copies. For
kpeter@710
    11
 * precise terms see the accompanying LICENSE file.
kpeter@710
    12
 *
kpeter@710
    13
 * This software is provided "AS IS" with no warranty of any kind,
kpeter@710
    14
 * express or implied, and with no claim as to its suitability for any
kpeter@710
    15
 * purpose.
kpeter@710
    16
 *
kpeter@710
    17
 */
kpeter@710
    18
kpeter@710
    19
namespace lemon {
kpeter@710
    20
kpeter@710
    21
/**
kpeter@710
    22
\page min_cost_flow Minimum Cost Flow Problem
kpeter@710
    23
kpeter@710
    24
\section mcf_def Definition (GEQ form)
kpeter@710
    25
kpeter@710
    26
The \e minimum \e cost \e flow \e problem is to find a feasible flow of
kpeter@710
    27
minimum total cost from a set of supply nodes to a set of demand nodes
kpeter@710
    28
in a network with capacity constraints (lower and upper bounds)
kpeter@802
    29
and arc costs \ref amo93networkflows.
kpeter@710
    30
kpeter@710
    31
Formally, let \f$G=(V,A)\f$ be a digraph, \f$lower: A\rightarrow\mathbf{R}\f$,
kpeter@710
    32
\f$upper: A\rightarrow\mathbf{R}\cup\{+\infty\}\f$ denote the lower and
kpeter@710
    33
upper bounds for the flow values on the arcs, for which
kpeter@710
    34
\f$lower(uv) \leq upper(uv)\f$ must hold for all \f$uv\in A\f$,
kpeter@710
    35
\f$cost: A\rightarrow\mathbf{R}\f$ denotes the cost per unit flow
kpeter@710
    36
on the arcs and \f$sup: V\rightarrow\mathbf{R}\f$ denotes the
kpeter@710
    37
signed supply values of the nodes.
kpeter@710
    38
If \f$sup(u)>0\f$, then \f$u\f$ is a supply node with \f$sup(u)\f$
kpeter@710
    39
supply, if \f$sup(u)<0\f$, then \f$u\f$ is a demand node with
kpeter@710
    40
\f$-sup(u)\f$ demand.
kpeter@710
    41
A minimum cost flow is an \f$f: A\rightarrow\mathbf{R}\f$ solution
kpeter@710
    42
of the following optimization problem.
kpeter@710
    43
kpeter@710
    44
\f[ \min\sum_{uv\in A} f(uv) \cdot cost(uv) \f]
kpeter@710
    45
\f[ \sum_{uv\in A} f(uv) - \sum_{vu\in A} f(vu) \geq
kpeter@710
    46
    sup(u) \quad \forall u\in V \f]
kpeter@710
    47
\f[ lower(uv) \leq f(uv) \leq upper(uv) \quad \forall uv\in A \f]
kpeter@710
    48
kpeter@710
    49
The sum of the supply values, i.e. \f$\sum_{u\in V} sup(u)\f$ must be
kpeter@710
    50
zero or negative in order to have a feasible solution (since the sum
kpeter@710
    51
of the expressions on the left-hand side of the inequalities is zero).
kpeter@710
    52
It means that the total demand must be greater or equal to the total
kpeter@710
    53
supply and all the supplies have to be carried out from the supply nodes,
kpeter@710
    54
but there could be demands that are not satisfied.
kpeter@710
    55
If \f$\sum_{u\in V} sup(u)\f$ is zero, then all the supply/demand
kpeter@710
    56
constraints have to be satisfied with equality, i.e. all demands
kpeter@710
    57
have to be satisfied and all supplies have to be used.
kpeter@710
    58
kpeter@710
    59
kpeter@710
    60
\section mcf_algs Algorithms
kpeter@710
    61
kpeter@710
    62
LEMON contains several algorithms for solving this problem, for more
kpeter@710
    63
information see \ref min_cost_flow_algs "Minimum Cost Flow Algorithms".
kpeter@710
    64
kpeter@710
    65
A feasible solution for this problem can be found using \ref Circulation.
kpeter@710
    66
kpeter@710
    67
kpeter@710
    68
\section mcf_dual Dual Solution
kpeter@710
    69
kpeter@710
    70
The dual solution of the minimum cost flow problem is represented by
kpeter@710
    71
node potentials \f$\pi: V\rightarrow\mathbf{R}\f$.
kpeter@710
    72
An \f$f: A\rightarrow\mathbf{R}\f$ primal feasible solution is optimal
kpeter@710
    73
if and only if for some \f$\pi: V\rightarrow\mathbf{R}\f$ node potentials
kpeter@710
    74
the following \e complementary \e slackness optimality conditions hold.
kpeter@710
    75
kpeter@710
    76
 - For all \f$uv\in A\f$ arcs:
kpeter@710
    77
   - if \f$cost^\pi(uv)>0\f$, then \f$f(uv)=lower(uv)\f$;
kpeter@710
    78
   - if \f$lower(uv)<f(uv)<upper(uv)\f$, then \f$cost^\pi(uv)=0\f$;
kpeter@710
    79
   - if \f$cost^\pi(uv)<0\f$, then \f$f(uv)=upper(uv)\f$.
kpeter@710
    80
 - For all \f$u\in V\f$ nodes:
kpeter@833
    81
   - \f$\pi(u)\leq 0\f$;
kpeter@710
    82
   - if \f$\sum_{uv\in A} f(uv) - \sum_{vu\in A} f(vu) \neq sup(u)\f$,
kpeter@710
    83
     then \f$\pi(u)=0\f$.
alpar@956
    84
kpeter@710
    85
Here \f$cost^\pi(uv)\f$ denotes the \e reduced \e cost of the arc
kpeter@710
    86
\f$uv\in A\f$ with respect to the potential function \f$\pi\f$, i.e.
kpeter@710
    87
\f[ cost^\pi(uv) = cost(uv) + \pi(u) - \pi(v).\f]
kpeter@710
    88
kpeter@710
    89
All algorithms provide dual solution (node potentials), as well,
kpeter@710
    90
if an optimal flow is found.
kpeter@710
    91
kpeter@710
    92
kpeter@710
    93
\section mcf_eq Equality Form
kpeter@710
    94
kpeter@710
    95
The above \ref mcf_def "definition" is actually more general than the
kpeter@710
    96
usual formulation of the minimum cost flow problem, in which strict
kpeter@710
    97
equalities are required in the supply/demand contraints.
kpeter@710
    98
kpeter@710
    99
\f[ \min\sum_{uv\in A} f(uv) \cdot cost(uv) \f]
kpeter@710
   100
\f[ \sum_{uv\in A} f(uv) - \sum_{vu\in A} f(vu) =
kpeter@710
   101
    sup(u) \quad \forall u\in V \f]
kpeter@710
   102
\f[ lower(uv) \leq f(uv) \leq upper(uv) \quad \forall uv\in A \f]
kpeter@710
   103
kpeter@710
   104
However if the sum of the supply values is zero, then these two problems
kpeter@710
   105
are equivalent.
kpeter@710
   106
The \ref min_cost_flow_algs "algorithms" in LEMON support the general
kpeter@710
   107
form, so if you need the equality form, you have to ensure this additional
kpeter@710
   108
contraint manually.
kpeter@710
   109
kpeter@710
   110
kpeter@710
   111
\section mcf_leq Opposite Inequalites (LEQ Form)
kpeter@710
   112
kpeter@710
   113
Another possible definition of the minimum cost flow problem is
kpeter@710
   114
when there are <em>"less or equal"</em> (LEQ) supply/demand constraints,
kpeter@710
   115
instead of the <em>"greater or equal"</em> (GEQ) constraints.
kpeter@710
   116
kpeter@710
   117
\f[ \min\sum_{uv\in A} f(uv) \cdot cost(uv) \f]
kpeter@710
   118
\f[ \sum_{uv\in A} f(uv) - \sum_{vu\in A} f(vu) \leq
kpeter@710
   119
    sup(u) \quad \forall u\in V \f]
kpeter@710
   120
\f[ lower(uv) \leq f(uv) \leq upper(uv) \quad \forall uv\in A \f]
kpeter@710
   121
alpar@956
   122
It means that the total demand must be less or equal to the
kpeter@710
   123
total supply (i.e. \f$\sum_{u\in V} sup(u)\f$ must be zero or
kpeter@710
   124
positive) and all the demands have to be satisfied, but there
kpeter@710
   125
could be supplies that are not carried out from the supply
kpeter@710
   126
nodes.
kpeter@710
   127
The equality form is also a special case of this form, of course.
kpeter@710
   128
kpeter@710
   129
You could easily transform this case to the \ref mcf_def "GEQ form"
kpeter@710
   130
of the problem by reversing the direction of the arcs and taking the
kpeter@710
   131
negative of the supply values (e.g. using \ref ReverseDigraph and
kpeter@710
   132
\ref NegMap adaptors).
kpeter@710
   133
However \ref NetworkSimplex algorithm also supports this form directly
kpeter@710
   134
for the sake of convenience.
kpeter@710
   135
kpeter@710
   136
Note that the optimality conditions for this supply constraint type are
kpeter@710
   137
slightly differ from the conditions that are discussed for the GEQ form,
kpeter@710
   138
namely the potentials have to be non-negative instead of non-positive.
kpeter@710
   139
An \f$f: A\rightarrow\mathbf{R}\f$ feasible solution of this problem
kpeter@710
   140
is optimal if and only if for some \f$\pi: V\rightarrow\mathbf{R}\f$
kpeter@710
   141
node potentials the following conditions hold.
kpeter@710
   142
kpeter@710
   143
 - For all \f$uv\in A\f$ arcs:
kpeter@710
   144
   - if \f$cost^\pi(uv)>0\f$, then \f$f(uv)=lower(uv)\f$;
kpeter@710
   145
   - if \f$lower(uv)<f(uv)<upper(uv)\f$, then \f$cost^\pi(uv)=0\f$;
kpeter@710
   146
   - if \f$cost^\pi(uv)<0\f$, then \f$f(uv)=upper(uv)\f$.
kpeter@710
   147
 - For all \f$u\in V\f$ nodes:
kpeter@833
   148
   - \f$\pi(u)\geq 0\f$;
kpeter@710
   149
   - if \f$\sum_{uv\in A} f(uv) - \sum_{vu\in A} f(vu) \neq sup(u)\f$,
kpeter@710
   150
     then \f$\pi(u)=0\f$.
kpeter@710
   151
kpeter@710
   152
*/
kpeter@710
   153
}