Changeset 609:e6927fe719e6 in lemonmain for doc/groups.dox
 Timestamp:
 04/17/09 18:04:36 (12 years ago)
 Branch:
 default
 Phase:
 public
 File:

 1 edited
Legend:
 Unmodified
 Added
 Removed

doc/groups.dox
r455 r609 318 318 The \e maximum \e flow \e problem is to find a flow of maximum value between 319 319 a single source and a single target. Formally, there is a \f$G=(V,A)\f$ 320 digraph, a \f$cap: A\rightarrow\mathbf{R}^+_0\f$ capacity function and320 digraph, a \f$cap: A\rightarrow\mathbf{R}^+_0\f$ capacity function and 321 321 \f$s, t \in V\f$ source and target nodes. 322 A maximum flow is an \f$f: A\rightarrow\mathbf{R}^+_0\f$ solution of the322 A maximum flow is an \f$f: A\rightarrow\mathbf{R}^+_0\f$ solution of the 323 323 following optimization problem. 324 324 325 \f[ \max\sum_{ a\in\delta_{out}(s)}f(a)  \sum_{a\in\delta_{in}(s)}f(a) \f]326 \f[ \sum_{ a\in\delta_{out}(v)} f(a) = \sum_{a\in\delta_{in}(v)} f(a)327 \q quad \forall v\in V\setminus\{s,t\} \f]328 \f[ 0 \leq f( a) \leq cap(a) \qquad \forall a\in A \f]325 \f[ \max\sum_{sv\in A} f(sv)  \sum_{vs\in A} f(vs) \f] 326 \f[ \sum_{uv\in A} f(uv) = \sum_{vu\in A} f(vu) 327 \quad \forall u\in V\setminus\{s,t\} \f] 328 \f[ 0 \leq f(uv) \leq cap(uv) \quad \forall uv\in A \f] 329 329 330 330 LEMON contains several algorithms for solving maximum flow problems: … … 346 346 \brief Algorithms for finding minimum cost flows and circulations. 347 347 348 This group describes the algorithms for finding minimum cost flows and348 This group contains the algorithms for finding minimum cost flows and 349 349 circulations. 350 350 351 351 The \e minimum \e cost \e flow \e problem is to find a feasible flow of 352 352 minimum total cost from a set of supply nodes to a set of demand nodes 353 in a network with capacity constraints and arc costs. 353 in a network with capacity constraints (lower and upper bounds) 354 and arc costs. 354 355 Formally, let \f$G=(V,A)\f$ be a digraph, 355 356 \f$lower, upper: A\rightarrow\mathbf{Z}^+_0\f$ denote the lower and 356 upper bounds for the flow values on the arcs, 357 upper bounds for the flow values on the arcs, for which 358 \f$0 \leq lower(uv) \leq upper(uv)\f$ holds for all \f$uv\in A\f$. 357 359 \f$cost: A\rightarrow\mathbf{Z}^+_0\f$ denotes the cost per unit flow 358 on the arcs, and 359 \f$supply: V\rightarrow\mathbf{Z}\f$ denotes the supply/demand values 360 of the nodes. 361 A minimum cost flow is an \f$f:A\rightarrow\mathbf{R}^+_0\f$ solution of 362 the following optimization problem. 363 364 \f[ \min\sum_{a\in A} f(a) cost(a) \f] 365 \f[ \sum_{a\in\delta_{out}(v)} f(a)  \sum_{a\in\delta_{in}(v)} f(a) = 366 supply(v) \qquad \forall v\in V \f] 367 \f[ lower(a) \leq f(a) \leq upper(a) \qquad \forall a\in A \f] 368 369 LEMON contains several algorithms for solving minimum cost flow problems: 370  \ref CycleCanceling Cyclecanceling algorithms. 371  \ref CapacityScaling Successive shortest path algorithm with optional 360 on the arcs, and \f$sup: V\rightarrow\mathbf{Z}\f$ denotes the 361 signed supply values of the nodes. 362 If \f$sup(u)>0\f$, then \f$u\f$ is a supply node with \f$sup(u)\f$ 363 supply, if \f$sup(u)<0\f$, then \f$u\f$ is a demand node with 364 \f$sup(u)\f$ demand. 365 A minimum cost flow is an \f$f: A\rightarrow\mathbf{Z}^+_0\f$ solution 366 of the following optimization problem. 367 368 \f[ \min\sum_{uv\in A} f(uv) \cdot cost(uv) \f] 369 \f[ \sum_{uv\in A} f(uv)  \sum_{vu\in A} f(vu) \geq 370 sup(u) \quad \forall u\in V \f] 371 \f[ lower(uv) \leq f(uv) \leq upper(uv) \quad \forall uv\in A \f] 372 373 The sum of the supply values, i.e. \f$\sum_{u\in V} sup(u)\f$ must be 374 zero or negative in order to have a feasible solution (since the sum 375 of the expressions on the lefthand side of the inequalities is zero). 376 It means that the total demand must be greater or equal to the total 377 supply and all the supplies have to be carried out from the supply nodes, 378 but there could be demands that are not satisfied. 379 If \f$\sum_{u\in V} sup(u)\f$ is zero, then all the supply/demand 380 constraints have to be satisfied with equality, i.e. all demands 381 have to be satisfied and all supplies have to be used. 382 383 If you need the opposite inequalities in the supply/demand constraints 384 (i.e. the total demand is less than the total supply and all the demands 385 have to be satisfied while there could be supplies that are not used), 386 then you could easily transform the problem to the above form by reversing 387 the direction of the arcs and taking the negative of the supply values 388 (e.g. using \ref ReverseDigraph and \ref NegMap adaptors). 389 However \ref NetworkSimplex algorithm also supports this form directly 390 for the sake of convenience. 391 392 A feasible solution for this problem can be found using \ref Circulation. 393 394 Note that the above formulation is actually more general than the usual 395 definition of the minimum cost flow problem, in which strict equalities 396 are required in the supply/demand contraints, i.e. 397 398 \f[ \sum_{uv\in A} f(uv)  \sum_{vu\in A} f(vu) = 399 sup(u) \quad \forall u\in V. \f] 400 401 However if the sum of the supply values is zero, then these two problems 402 are equivalent. So if you need the equality form, you have to ensure this 403 additional contraint for the algorithms. 404 405 The dual solution of the minimum cost flow problem is represented by node 406 potentials \f$\pi: V\rightarrow\mathbf{Z}\f$. 407 An \f$f: A\rightarrow\mathbf{Z}^+_0\f$ feasible solution of the problem 408 is optimal if and only if for some \f$\pi: V\rightarrow\mathbf{Z}\f$ 409 node potentials the following \e complementary \e slackness optimality 410 conditions hold. 411 412  For all \f$uv\in A\f$ arcs: 413  if \f$cost^\pi(uv)>0\f$, then \f$f(uv)=lower(uv)\f$; 414  if \f$lower(uv)<f(uv)<upper(uv)\f$, then \f$cost^\pi(uv)=0\f$; 415  if \f$cost^\pi(uv)<0\f$, then \f$f(uv)=upper(uv)\f$. 416  For all \f$u\in V\f$: 417  if \f$\sum_{uv\in A} f(uv)  \sum_{vu\in A} f(vu) \neq sup(u)\f$, 418 then \f$\pi(u)=0\f$. 419 420 Here \f$cost^\pi(uv)\f$ denotes the \e reduced \e cost of the arc 421 \f$uv\in A\f$ with respect to the node potentials \f$\pi\f$, i.e. 422 \f[ cost^\pi(uv) = cost(uv) + \pi(u)  \pi(v).\f] 423 424 All algorithms provide dual solution (node potentials) as well 425 if an optimal flow is found. 426 427 LEMON contains several algorithms for solving minimum cost flow problems. 428  \ref NetworkSimplex Primal Network Simplex algorithm with various 429 pivot strategies. 430  \ref CostScaling PushRelabel and AugmentRelabel algorithms based on 431 cost scaling. 432  \ref CapacityScaling Successive Shortest %Path algorithm with optional 372 433 capacity scaling. 373  \ref CostScaling Pushrelabel and augmentrelabel algorithms based on 374 cost scaling. 375  \ref NetworkSimplex Primal network simplex algorithm with various 376 pivot strategies. 434  \ref CancelAndTighten The Cancel and Tighten algorithm. 435  \ref CycleCanceling CycleCanceling algorithms. 436 437 Most of these implementations support the general inequality form of the 438 minimum cost flow problem, but CancelAndTighten and CycleCanceling 439 only support the equality form due to the primal method they use. 440 441 In general NetworkSimplex is the most efficient implementation, 442 but in special cases other algorithms could be faster. 443 For example, if the total supply and/or capacities are rather small, 444 CapacityScaling is usually the fastest algorithm (without effective scaling). 377 445 */ 378 446
Note: See TracChangeset
for help on using the changeset viewer.