0
2
0
1 | 1 |
/* -*- mode: C++; indent-tabs-mode: nil; -*- |
2 | 2 |
* |
3 | 3 |
* This file is a part of LEMON, a generic C++ optimization library. |
4 | 4 |
* |
5 | 5 |
* Copyright (C) 2003-2009 |
6 | 6 |
* Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport |
7 | 7 |
* (Egervary Research Group on Combinatorial Optimization, EGRES). |
8 | 8 |
* |
9 | 9 |
* Permission to use, modify and distribute this software is granted |
10 | 10 |
* provided that this copyright notice appears in all copies. For |
11 | 11 |
* precise terms see the accompanying LICENSE file. |
12 | 12 |
* |
13 | 13 |
* This software is provided "AS IS" with no warranty of any kind, |
14 | 14 |
* express or implied, and with no claim as to its suitability for any |
15 | 15 |
* purpose. |
16 | 16 |
* |
17 | 17 |
*/ |
18 | 18 |
|
19 | 19 |
#ifndef LEMON_NETWORK_SIMPLEX_H |
20 | 20 |
#define LEMON_NETWORK_SIMPLEX_H |
21 | 21 |
|
22 | 22 |
/// \ingroup min_cost_flow |
23 | 23 |
/// |
24 | 24 |
/// \file |
25 | 25 |
/// \brief Network Simplex algorithm for finding a minimum cost flow. |
26 | 26 |
|
27 | 27 |
#include <vector> |
28 | 28 |
#include <limits> |
29 | 29 |
#include <algorithm> |
30 | 30 |
|
31 | 31 |
#include <lemon/core.h> |
32 | 32 |
#include <lemon/math.h> |
33 | 33 |
|
34 | 34 |
namespace lemon { |
35 | 35 |
|
36 | 36 |
/// \addtogroup min_cost_flow |
37 | 37 |
/// @{ |
38 | 38 |
|
39 | 39 |
/// \brief Implementation of the primal Network Simplex algorithm |
40 | 40 |
/// for finding a \ref min_cost_flow "minimum cost flow". |
41 | 41 |
/// |
42 | 42 |
/// \ref NetworkSimplex implements the primal Network Simplex algorithm |
43 | 43 |
/// for finding a \ref min_cost_flow "minimum cost flow". |
44 | 44 |
/// This algorithm is a specialized version of the linear programming |
45 | 45 |
/// simplex method directly for the minimum cost flow problem. |
46 | 46 |
/// It is one of the most efficient solution methods. |
47 | 47 |
/// |
48 | 48 |
/// In general this class is the fastest implementation available |
49 | 49 |
/// in LEMON for the minimum cost flow problem. |
50 | 50 |
/// |
51 | 51 |
/// \tparam GR The digraph type the algorithm runs on. |
52 |
/// \tparam V The value type used in the algorithm. |
|
53 |
/// By default it is \c int. |
|
52 |
/// \tparam F The value type used for flow amounts, capacity bounds |
|
53 |
/// and supply values in the algorithm. By default it is \c int. |
|
54 |
/// \tparam C The value type used for costs and potentials in the |
|
55 |
/// algorithm. By default it is the same as \c F. |
|
54 | 56 |
/// |
55 |
/// \warning |
|
57 |
/// \warning Both value types must be signed integer types. |
|
56 | 58 |
/// |
57 | 59 |
/// \note %NetworkSimplex provides five different pivot rule |
58 | 60 |
/// implementations. For more information see \ref PivotRule. |
59 |
template <typename GR, typename |
|
61 |
template <typename GR, typename F = int, typename C = F> |
|
60 | 62 |
class NetworkSimplex |
61 | 63 |
{ |
62 | 64 |
public: |
63 | 65 |
|
64 |
/// The value type of the algorithm |
|
65 |
typedef V Value; |
|
66 |
/// The flow type of the algorithm |
|
67 |
typedef F Flow; |
|
68 |
/// The cost type of the algorithm |
|
69 |
typedef C Cost; |
|
66 | 70 |
/// The type of the flow map |
67 |
typedef typename GR::template ArcMap< |
|
71 |
typedef typename GR::template ArcMap<Flow> FlowMap; |
|
68 | 72 |
/// The type of the potential map |
69 |
typedef typename GR::template NodeMap< |
|
73 |
typedef typename GR::template NodeMap<Cost> PotentialMap; |
|
70 | 74 |
|
71 | 75 |
public: |
72 | 76 |
|
73 | 77 |
/// \brief Enum type for selecting the pivot rule. |
74 | 78 |
/// |
75 | 79 |
/// Enum type for selecting the pivot rule for the \ref run() |
76 | 80 |
/// function. |
77 | 81 |
/// |
78 | 82 |
/// \ref NetworkSimplex provides five different pivot rule |
79 | 83 |
/// implementations that significantly affect the running time |
80 | 84 |
/// of the algorithm. |
81 | 85 |
/// By default \ref BLOCK_SEARCH "Block Search" is used, which |
82 | 86 |
/// proved to be the most efficient and the most robust on various |
83 | 87 |
/// test inputs according to our benchmark tests. |
84 | 88 |
/// However another pivot rule can be selected using the \ref run() |
85 | 89 |
/// function with the proper parameter. |
86 | 90 |
enum PivotRule { |
87 | 91 |
|
88 | 92 |
/// The First Eligible pivot rule. |
89 | 93 |
/// The next eligible arc is selected in a wraparound fashion |
90 | 94 |
/// in every iteration. |
91 | 95 |
FIRST_ELIGIBLE, |
92 | 96 |
|
93 | 97 |
/// The Best Eligible pivot rule. |
94 | 98 |
/// The best eligible arc is selected in every iteration. |
95 | 99 |
BEST_ELIGIBLE, |
96 | 100 |
|
97 | 101 |
/// The Block Search pivot rule. |
98 | 102 |
/// A specified number of arcs are examined in every iteration |
99 | 103 |
/// in a wraparound fashion and the best eligible arc is selected |
100 | 104 |
/// from this block. |
101 | 105 |
BLOCK_SEARCH, |
102 | 106 |
|
103 | 107 |
/// The Candidate List pivot rule. |
104 | 108 |
/// In a major iteration a candidate list is built from eligible arcs |
105 | 109 |
/// in a wraparound fashion and in the following minor iterations |
106 | 110 |
/// the best eligible arc is selected from this list. |
107 | 111 |
CANDIDATE_LIST, |
108 | 112 |
|
109 | 113 |
/// The Altering Candidate List pivot rule. |
110 | 114 |
/// It is a modified version of the Candidate List method. |
111 | 115 |
/// It keeps only the several best eligible arcs from the former |
112 | 116 |
/// candidate list and extends this list in every iteration. |
113 | 117 |
ALTERING_LIST |
114 | 118 |
}; |
115 | 119 |
|
116 | 120 |
private: |
117 | 121 |
|
118 | 122 |
TEMPLATE_DIGRAPH_TYPEDEFS(GR); |
119 | 123 |
|
120 |
typedef typename GR::template ArcMap<Value> ValueArcMap; |
|
121 |
typedef typename GR::template NodeMap<Value> ValueNodeMap; |
|
124 |
typedef typename GR::template ArcMap<Flow> FlowArcMap; |
|
125 |
typedef typename GR::template ArcMap<Cost> CostArcMap; |
|
126 |
typedef typename GR::template NodeMap<Flow> FlowNodeMap; |
|
122 | 127 |
|
123 | 128 |
typedef std::vector<Arc> ArcVector; |
124 | 129 |
typedef std::vector<Node> NodeVector; |
125 | 130 |
typedef std::vector<int> IntVector; |
126 | 131 |
typedef std::vector<bool> BoolVector; |
127 |
typedef std::vector< |
|
132 |
typedef std::vector<Flow> FlowVector; |
|
133 |
typedef std::vector<Cost> CostVector; |
|
128 | 134 |
|
129 | 135 |
// State constants for arcs |
130 | 136 |
enum ArcStateEnum { |
131 | 137 |
STATE_UPPER = -1, |
132 | 138 |
STATE_TREE = 0, |
133 | 139 |
STATE_LOWER = 1 |
134 | 140 |
}; |
135 | 141 |
|
136 | 142 |
private: |
137 | 143 |
|
138 | 144 |
// Data related to the underlying digraph |
139 | 145 |
const GR &_graph; |
140 | 146 |
int _node_num; |
141 | 147 |
int _arc_num; |
142 | 148 |
|
143 | 149 |
// Parameters of the problem |
144 |
ValueArcMap *_plower; |
|
145 |
ValueArcMap *_pupper; |
|
146 |
ValueArcMap *_pcost; |
|
147 |
ValueNodeMap *_psupply; |
|
150 |
FlowArcMap *_plower; |
|
151 |
FlowArcMap *_pupper; |
|
152 |
CostArcMap *_pcost; |
|
153 |
FlowNodeMap *_psupply; |
|
148 | 154 |
bool _pstsup; |
149 | 155 |
Node _psource, _ptarget; |
150 |
|
|
156 |
Flow _pstflow; |
|
151 | 157 |
|
152 | 158 |
// Result maps |
153 | 159 |
FlowMap *_flow_map; |
154 | 160 |
PotentialMap *_potential_map; |
155 | 161 |
bool _local_flow; |
156 | 162 |
bool _local_potential; |
157 | 163 |
|
158 | 164 |
// Data structures for storing the digraph |
159 | 165 |
IntNodeMap _node_id; |
160 | 166 |
ArcVector _arc_ref; |
161 | 167 |
IntVector _source; |
162 | 168 |
IntVector _target; |
163 | 169 |
|
164 | 170 |
// Node and arc data |
165 |
ValueVector _cap; |
|
166 |
ValueVector _cost; |
|
167 |
ValueVector _supply; |
|
168 |
ValueVector _flow; |
|
169 |
|
|
171 |
FlowVector _cap; |
|
172 |
CostVector _cost; |
|
173 |
FlowVector _supply; |
|
174 |
FlowVector _flow; |
|
175 |
CostVector _pi; |
|
170 | 176 |
|
171 | 177 |
// Data for storing the spanning tree structure |
172 | 178 |
IntVector _parent; |
173 | 179 |
IntVector _pred; |
174 | 180 |
IntVector _thread; |
175 | 181 |
IntVector _rev_thread; |
176 | 182 |
IntVector _succ_num; |
177 | 183 |
IntVector _last_succ; |
178 | 184 |
IntVector _dirty_revs; |
179 | 185 |
BoolVector _forward; |
180 | 186 |
IntVector _state; |
181 | 187 |
int _root; |
182 | 188 |
|
183 | 189 |
// Temporary data used in the current pivot iteration |
184 | 190 |
int in_arc, join, u_in, v_in, u_out, v_out; |
185 | 191 |
int first, second, right, last; |
186 | 192 |
int stem, par_stem, new_stem; |
187 |
|
|
193 |
Flow delta; |
|
188 | 194 |
|
189 | 195 |
private: |
190 | 196 |
|
191 | 197 |
// Implementation of the First Eligible pivot rule |
192 | 198 |
class FirstEligiblePivotRule |
193 | 199 |
{ |
194 | 200 |
private: |
195 | 201 |
|
196 | 202 |
// References to the NetworkSimplex class |
197 | 203 |
const IntVector &_source; |
198 | 204 |
const IntVector &_target; |
199 |
const |
|
205 |
const CostVector &_cost; |
|
200 | 206 |
const IntVector &_state; |
201 |
const |
|
207 |
const CostVector &_pi; |
|
202 | 208 |
int &_in_arc; |
203 | 209 |
int _arc_num; |
204 | 210 |
|
205 | 211 |
// Pivot rule data |
206 | 212 |
int _next_arc; |
207 | 213 |
|
208 | 214 |
public: |
209 | 215 |
|
210 | 216 |
// Constructor |
211 | 217 |
FirstEligiblePivotRule(NetworkSimplex &ns) : |
212 | 218 |
_source(ns._source), _target(ns._target), |
213 | 219 |
_cost(ns._cost), _state(ns._state), _pi(ns._pi), |
214 | 220 |
_in_arc(ns.in_arc), _arc_num(ns._arc_num), _next_arc(0) |
215 | 221 |
{} |
216 | 222 |
|
217 | 223 |
// Find next entering arc |
218 | 224 |
bool findEnteringArc() { |
219 |
|
|
225 |
Cost c; |
|
220 | 226 |
for (int e = _next_arc; e < _arc_num; ++e) { |
221 | 227 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
222 | 228 |
if (c < 0) { |
223 | 229 |
_in_arc = e; |
224 | 230 |
_next_arc = e + 1; |
225 | 231 |
return true; |
226 | 232 |
} |
227 | 233 |
} |
228 | 234 |
for (int e = 0; e < _next_arc; ++e) { |
229 | 235 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
230 | 236 |
if (c < 0) { |
231 | 237 |
_in_arc = e; |
232 | 238 |
_next_arc = e + 1; |
233 | 239 |
return true; |
234 | 240 |
} |
235 | 241 |
} |
236 | 242 |
return false; |
237 | 243 |
} |
238 | 244 |
|
239 | 245 |
}; //class FirstEligiblePivotRule |
240 | 246 |
|
241 | 247 |
|
242 | 248 |
// Implementation of the Best Eligible pivot rule |
243 | 249 |
class BestEligiblePivotRule |
244 | 250 |
{ |
245 | 251 |
private: |
246 | 252 |
|
247 | 253 |
// References to the NetworkSimplex class |
248 | 254 |
const IntVector &_source; |
249 | 255 |
const IntVector &_target; |
250 |
const |
|
256 |
const CostVector &_cost; |
|
251 | 257 |
const IntVector &_state; |
252 |
const |
|
258 |
const CostVector &_pi; |
|
253 | 259 |
int &_in_arc; |
254 | 260 |
int _arc_num; |
255 | 261 |
|
256 | 262 |
public: |
257 | 263 |
|
258 | 264 |
// Constructor |
259 | 265 |
BestEligiblePivotRule(NetworkSimplex &ns) : |
260 | 266 |
_source(ns._source), _target(ns._target), |
261 | 267 |
_cost(ns._cost), _state(ns._state), _pi(ns._pi), |
262 | 268 |
_in_arc(ns.in_arc), _arc_num(ns._arc_num) |
263 | 269 |
{} |
264 | 270 |
|
265 | 271 |
// Find next entering arc |
266 | 272 |
bool findEnteringArc() { |
267 |
|
|
273 |
Cost c, min = 0; |
|
268 | 274 |
for (int e = 0; e < _arc_num; ++e) { |
269 | 275 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
270 | 276 |
if (c < min) { |
271 | 277 |
min = c; |
272 | 278 |
_in_arc = e; |
273 | 279 |
} |
274 | 280 |
} |
275 | 281 |
return min < 0; |
276 | 282 |
} |
277 | 283 |
|
278 | 284 |
}; //class BestEligiblePivotRule |
279 | 285 |
|
280 | 286 |
|
281 | 287 |
// Implementation of the Block Search pivot rule |
282 | 288 |
class BlockSearchPivotRule |
283 | 289 |
{ |
284 | 290 |
private: |
285 | 291 |
|
286 | 292 |
// References to the NetworkSimplex class |
287 | 293 |
const IntVector &_source; |
288 | 294 |
const IntVector &_target; |
289 |
const |
|
295 |
const CostVector &_cost; |
|
290 | 296 |
const IntVector &_state; |
291 |
const |
|
297 |
const CostVector &_pi; |
|
292 | 298 |
int &_in_arc; |
293 | 299 |
int _arc_num; |
294 | 300 |
|
295 | 301 |
// Pivot rule data |
296 | 302 |
int _block_size; |
297 | 303 |
int _next_arc; |
298 | 304 |
|
299 | 305 |
public: |
300 | 306 |
|
301 | 307 |
// Constructor |
302 | 308 |
BlockSearchPivotRule(NetworkSimplex &ns) : |
303 | 309 |
_source(ns._source), _target(ns._target), |
304 | 310 |
_cost(ns._cost), _state(ns._state), _pi(ns._pi), |
305 | 311 |
_in_arc(ns.in_arc), _arc_num(ns._arc_num), _next_arc(0) |
306 | 312 |
{ |
307 | 313 |
// The main parameters of the pivot rule |
308 | 314 |
const double BLOCK_SIZE_FACTOR = 2.0; |
309 | 315 |
const int MIN_BLOCK_SIZE = 10; |
310 | 316 |
|
311 | 317 |
_block_size = std::max( int(BLOCK_SIZE_FACTOR * sqrt(_arc_num)), |
312 | 318 |
MIN_BLOCK_SIZE ); |
313 | 319 |
} |
314 | 320 |
|
315 | 321 |
// Find next entering arc |
316 | 322 |
bool findEnteringArc() { |
317 |
|
|
323 |
Cost c, min = 0; |
|
318 | 324 |
int cnt = _block_size; |
319 | 325 |
int e, min_arc = _next_arc; |
320 | 326 |
for (e = _next_arc; e < _arc_num; ++e) { |
321 | 327 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
322 | 328 |
if (c < min) { |
323 | 329 |
min = c; |
324 | 330 |
min_arc = e; |
325 | 331 |
} |
326 | 332 |
if (--cnt == 0) { |
327 | 333 |
if (min < 0) break; |
328 | 334 |
cnt = _block_size; |
329 | 335 |
} |
330 | 336 |
} |
331 | 337 |
if (min == 0 || cnt > 0) { |
332 | 338 |
for (e = 0; e < _next_arc; ++e) { |
333 | 339 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
334 | 340 |
if (c < min) { |
335 | 341 |
min = c; |
336 | 342 |
min_arc = e; |
337 | 343 |
} |
338 | 344 |
if (--cnt == 0) { |
339 | 345 |
if (min < 0) break; |
340 | 346 |
cnt = _block_size; |
341 | 347 |
} |
342 | 348 |
} |
343 | 349 |
} |
344 | 350 |
if (min >= 0) return false; |
345 | 351 |
_in_arc = min_arc; |
346 | 352 |
_next_arc = e; |
347 | 353 |
return true; |
348 | 354 |
} |
349 | 355 |
|
350 | 356 |
}; //class BlockSearchPivotRule |
351 | 357 |
|
352 | 358 |
|
353 | 359 |
// Implementation of the Candidate List pivot rule |
354 | 360 |
class CandidateListPivotRule |
355 | 361 |
{ |
356 | 362 |
private: |
357 | 363 |
|
358 | 364 |
// References to the NetworkSimplex class |
359 | 365 |
const IntVector &_source; |
360 | 366 |
const IntVector &_target; |
361 |
const |
|
367 |
const CostVector &_cost; |
|
362 | 368 |
const IntVector &_state; |
363 |
const |
|
369 |
const CostVector &_pi; |
|
364 | 370 |
int &_in_arc; |
365 | 371 |
int _arc_num; |
366 | 372 |
|
367 | 373 |
// Pivot rule data |
368 | 374 |
IntVector _candidates; |
369 | 375 |
int _list_length, _minor_limit; |
370 | 376 |
int _curr_length, _minor_count; |
371 | 377 |
int _next_arc; |
372 | 378 |
|
373 | 379 |
public: |
374 | 380 |
|
375 | 381 |
/// Constructor |
376 | 382 |
CandidateListPivotRule(NetworkSimplex &ns) : |
377 | 383 |
_source(ns._source), _target(ns._target), |
378 | 384 |
_cost(ns._cost), _state(ns._state), _pi(ns._pi), |
379 | 385 |
_in_arc(ns.in_arc), _arc_num(ns._arc_num), _next_arc(0) |
380 | 386 |
{ |
381 | 387 |
// The main parameters of the pivot rule |
382 | 388 |
const double LIST_LENGTH_FACTOR = 1.0; |
383 | 389 |
const int MIN_LIST_LENGTH = 10; |
384 | 390 |
const double MINOR_LIMIT_FACTOR = 0.1; |
385 | 391 |
const int MIN_MINOR_LIMIT = 3; |
386 | 392 |
|
387 | 393 |
_list_length = std::max( int(LIST_LENGTH_FACTOR * sqrt(_arc_num)), |
388 | 394 |
MIN_LIST_LENGTH ); |
389 | 395 |
_minor_limit = std::max( int(MINOR_LIMIT_FACTOR * _list_length), |
390 | 396 |
MIN_MINOR_LIMIT ); |
391 | 397 |
_curr_length = _minor_count = 0; |
392 | 398 |
_candidates.resize(_list_length); |
393 | 399 |
} |
394 | 400 |
|
395 | 401 |
/// Find next entering arc |
396 | 402 |
bool findEnteringArc() { |
397 |
|
|
403 |
Cost min, c; |
|
398 | 404 |
int e, min_arc = _next_arc; |
399 | 405 |
if (_curr_length > 0 && _minor_count < _minor_limit) { |
400 | 406 |
// Minor iteration: select the best eligible arc from the |
401 | 407 |
// current candidate list |
402 | 408 |
++_minor_count; |
403 | 409 |
min = 0; |
404 | 410 |
for (int i = 0; i < _curr_length; ++i) { |
405 | 411 |
e = _candidates[i]; |
406 | 412 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
407 | 413 |
if (c < min) { |
408 | 414 |
min = c; |
409 | 415 |
min_arc = e; |
410 | 416 |
} |
411 | 417 |
if (c >= 0) { |
412 | 418 |
_candidates[i--] = _candidates[--_curr_length]; |
413 | 419 |
} |
414 | 420 |
} |
415 | 421 |
if (min < 0) { |
416 | 422 |
_in_arc = min_arc; |
417 | 423 |
return true; |
418 | 424 |
} |
419 | 425 |
} |
420 | 426 |
|
421 | 427 |
// Major iteration: build a new candidate list |
422 | 428 |
min = 0; |
423 | 429 |
_curr_length = 0; |
424 | 430 |
for (e = _next_arc; e < _arc_num; ++e) { |
425 | 431 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
426 | 432 |
if (c < 0) { |
427 | 433 |
_candidates[_curr_length++] = e; |
428 | 434 |
if (c < min) { |
429 | 435 |
min = c; |
430 | 436 |
min_arc = e; |
431 | 437 |
} |
432 | 438 |
if (_curr_length == _list_length) break; |
433 | 439 |
} |
434 | 440 |
} |
435 | 441 |
if (_curr_length < _list_length) { |
436 | 442 |
for (e = 0; e < _next_arc; ++e) { |
437 | 443 |
c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
438 | 444 |
if (c < 0) { |
439 | 445 |
_candidates[_curr_length++] = e; |
440 | 446 |
if (c < min) { |
441 | 447 |
min = c; |
442 | 448 |
min_arc = e; |
443 | 449 |
} |
444 | 450 |
if (_curr_length == _list_length) break; |
445 | 451 |
} |
446 | 452 |
} |
447 | 453 |
} |
448 | 454 |
if (_curr_length == 0) return false; |
449 | 455 |
_minor_count = 1; |
450 | 456 |
_in_arc = min_arc; |
451 | 457 |
_next_arc = e; |
452 | 458 |
return true; |
453 | 459 |
} |
454 | 460 |
|
455 | 461 |
}; //class CandidateListPivotRule |
456 | 462 |
|
457 | 463 |
|
458 | 464 |
// Implementation of the Altering Candidate List pivot rule |
459 | 465 |
class AlteringListPivotRule |
460 | 466 |
{ |
461 | 467 |
private: |
462 | 468 |
|
463 | 469 |
// References to the NetworkSimplex class |
464 | 470 |
const IntVector &_source; |
465 | 471 |
const IntVector &_target; |
466 |
const |
|
472 |
const CostVector &_cost; |
|
467 | 473 |
const IntVector &_state; |
468 |
const |
|
474 |
const CostVector &_pi; |
|
469 | 475 |
int &_in_arc; |
470 | 476 |
int _arc_num; |
471 | 477 |
|
472 | 478 |
// Pivot rule data |
473 | 479 |
int _block_size, _head_length, _curr_length; |
474 | 480 |
int _next_arc; |
475 | 481 |
IntVector _candidates; |
476 |
|
|
482 |
CostVector _cand_cost; |
|
477 | 483 |
|
478 | 484 |
// Functor class to compare arcs during sort of the candidate list |
479 | 485 |
class SortFunc |
480 | 486 |
{ |
481 | 487 |
private: |
482 |
const |
|
488 |
const CostVector &_map; |
|
483 | 489 |
public: |
484 |
SortFunc(const |
|
490 |
SortFunc(const CostVector &map) : _map(map) {} |
|
485 | 491 |
bool operator()(int left, int right) { |
486 | 492 |
return _map[left] > _map[right]; |
487 | 493 |
} |
488 | 494 |
}; |
489 | 495 |
|
490 | 496 |
SortFunc _sort_func; |
491 | 497 |
|
492 | 498 |
public: |
493 | 499 |
|
494 | 500 |
// Constructor |
495 | 501 |
AlteringListPivotRule(NetworkSimplex &ns) : |
496 | 502 |
_source(ns._source), _target(ns._target), |
497 | 503 |
_cost(ns._cost), _state(ns._state), _pi(ns._pi), |
498 | 504 |
_in_arc(ns.in_arc), _arc_num(ns._arc_num), |
499 | 505 |
_next_arc(0), _cand_cost(ns._arc_num), _sort_func(_cand_cost) |
500 | 506 |
{ |
501 | 507 |
// The main parameters of the pivot rule |
502 | 508 |
const double BLOCK_SIZE_FACTOR = 1.5; |
503 | 509 |
const int MIN_BLOCK_SIZE = 10; |
504 | 510 |
const double HEAD_LENGTH_FACTOR = 0.1; |
505 | 511 |
const int MIN_HEAD_LENGTH = 3; |
506 | 512 |
|
507 | 513 |
_block_size = std::max( int(BLOCK_SIZE_FACTOR * sqrt(_arc_num)), |
508 | 514 |
MIN_BLOCK_SIZE ); |
509 | 515 |
_head_length = std::max( int(HEAD_LENGTH_FACTOR * _block_size), |
510 | 516 |
MIN_HEAD_LENGTH ); |
511 | 517 |
_candidates.resize(_head_length + _block_size); |
512 | 518 |
_curr_length = 0; |
513 | 519 |
} |
514 | 520 |
|
515 | 521 |
// Find next entering arc |
516 | 522 |
bool findEnteringArc() { |
517 | 523 |
// Check the current candidate list |
518 | 524 |
int e; |
519 | 525 |
for (int i = 0; i < _curr_length; ++i) { |
520 | 526 |
e = _candidates[i]; |
521 | 527 |
_cand_cost[e] = _state[e] * |
522 | 528 |
(_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
523 | 529 |
if (_cand_cost[e] >= 0) { |
524 | 530 |
_candidates[i--] = _candidates[--_curr_length]; |
525 | 531 |
} |
526 | 532 |
} |
527 | 533 |
|
528 | 534 |
// Extend the list |
529 | 535 |
int cnt = _block_size; |
530 | 536 |
int last_arc = 0; |
531 | 537 |
int limit = _head_length; |
532 | 538 |
|
533 | 539 |
for (int e = _next_arc; e < _arc_num; ++e) { |
534 | 540 |
_cand_cost[e] = _state[e] * |
535 | 541 |
(_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
536 | 542 |
if (_cand_cost[e] < 0) { |
537 | 543 |
_candidates[_curr_length++] = e; |
538 | 544 |
last_arc = e; |
539 | 545 |
} |
540 | 546 |
if (--cnt == 0) { |
541 | 547 |
if (_curr_length > limit) break; |
542 | 548 |
limit = 0; |
543 | 549 |
cnt = _block_size; |
544 | 550 |
} |
545 | 551 |
} |
546 | 552 |
if (_curr_length <= limit) { |
547 | 553 |
for (int e = 0; e < _next_arc; ++e) { |
548 | 554 |
_cand_cost[e] = _state[e] * |
549 | 555 |
(_cost[e] + _pi[_source[e]] - _pi[_target[e]]); |
550 | 556 |
if (_cand_cost[e] < 0) { |
551 | 557 |
_candidates[_curr_length++] = e; |
552 | 558 |
last_arc = e; |
553 | 559 |
} |
554 | 560 |
if (--cnt == 0) { |
555 | 561 |
if (_curr_length > limit) break; |
556 | 562 |
limit = 0; |
557 | 563 |
cnt = _block_size; |
558 | 564 |
} |
559 | 565 |
} |
560 | 566 |
} |
561 | 567 |
if (_curr_length == 0) return false; |
562 | 568 |
_next_arc = last_arc + 1; |
563 | 569 |
|
564 | 570 |
// Make heap of the candidate list (approximating a partial sort) |
565 | 571 |
make_heap( _candidates.begin(), _candidates.begin() + _curr_length, |
566 | 572 |
_sort_func ); |
567 | 573 |
|
568 | 574 |
// Pop the first element of the heap |
569 | 575 |
_in_arc = _candidates[0]; |
570 | 576 |
pop_heap( _candidates.begin(), _candidates.begin() + _curr_length, |
571 | 577 |
_sort_func ); |
572 | 578 |
_curr_length = std::min(_head_length, _curr_length - 1); |
573 | 579 |
return true; |
574 | 580 |
} |
575 | 581 |
|
576 | 582 |
}; //class AlteringListPivotRule |
577 | 583 |
|
578 | 584 |
public: |
579 | 585 |
|
580 | 586 |
/// \brief Constructor. |
581 | 587 |
/// |
582 | 588 |
/// Constructor. |
583 | 589 |
/// |
584 | 590 |
/// \param graph The digraph the algorithm runs on. |
585 | 591 |
NetworkSimplex(const GR& graph) : |
586 | 592 |
_graph(graph), |
587 | 593 |
_plower(NULL), _pupper(NULL), _pcost(NULL), |
588 | 594 |
_psupply(NULL), _pstsup(false), |
589 | 595 |
_flow_map(NULL), _potential_map(NULL), |
590 | 596 |
_local_flow(false), _local_potential(false), |
591 | 597 |
_node_id(graph) |
592 | 598 |
{ |
593 |
LEMON_ASSERT(std::numeric_limits<Value>::is_integer && |
|
594 |
std::numeric_limits<Value>::is_signed, |
|
595 |
|
|
599 |
LEMON_ASSERT(std::numeric_limits<Flow>::is_integer && |
|
600 |
std::numeric_limits<Flow>::is_signed, |
|
601 |
"The flow type of NetworkSimplex must be signed integer"); |
|
602 |
LEMON_ASSERT(std::numeric_limits<Cost>::is_integer && |
|
603 |
std::numeric_limits<Cost>::is_signed, |
|
604 |
"The cost type of NetworkSimplex must be signed integer"); |
|
596 | 605 |
} |
597 | 606 |
|
598 | 607 |
/// Destructor. |
599 | 608 |
~NetworkSimplex() { |
600 | 609 |
if (_local_flow) delete _flow_map; |
601 | 610 |
if (_local_potential) delete _potential_map; |
602 | 611 |
} |
603 | 612 |
|
604 | 613 |
/// \brief Set the lower bounds on the arcs. |
605 | 614 |
/// |
606 | 615 |
/// This function sets the lower bounds on the arcs. |
607 | 616 |
/// If neither this function nor \ref boundMaps() is used before |
608 | 617 |
/// calling \ref run(), the lower bounds will be set to zero |
609 | 618 |
/// on all arcs. |
610 | 619 |
/// |
611 | 620 |
/// \param map An arc map storing the lower bounds. |
612 |
/// Its \c Value type must be convertible to the \c |
|
621 |
/// Its \c Value type must be convertible to the \c Flow type |
|
613 | 622 |
/// of the algorithm. |
614 | 623 |
/// |
615 | 624 |
/// \return <tt>(*this)</tt> |
616 | 625 |
template <typename LOWER> |
617 | 626 |
NetworkSimplex& lowerMap(const LOWER& map) { |
618 | 627 |
delete _plower; |
619 |
_plower = new |
|
628 |
_plower = new FlowArcMap(_graph); |
|
620 | 629 |
for (ArcIt a(_graph); a != INVALID; ++a) { |
621 | 630 |
(*_plower)[a] = map[a]; |
622 | 631 |
} |
623 | 632 |
return *this; |
624 | 633 |
} |
625 | 634 |
|
626 | 635 |
/// \brief Set the upper bounds (capacities) on the arcs. |
627 | 636 |
/// |
628 | 637 |
/// This function sets the upper bounds (capacities) on the arcs. |
629 | 638 |
/// If none of the functions \ref upperMap(), \ref capacityMap() |
630 | 639 |
/// and \ref boundMaps() is used before calling \ref run(), |
631 | 640 |
/// the upper bounds (capacities) will be set to |
632 |
/// \c std::numeric_limits< |
|
641 |
/// \c std::numeric_limits<Flow>::max() on all arcs. |
|
633 | 642 |
/// |
634 | 643 |
/// \param map An arc map storing the upper bounds. |
635 |
/// Its \c Value type must be convertible to the \c |
|
644 |
/// Its \c Value type must be convertible to the \c Flow type |
|
636 | 645 |
/// of the algorithm. |
637 | 646 |
/// |
638 | 647 |
/// \return <tt>(*this)</tt> |
639 | 648 |
template<typename UPPER> |
640 | 649 |
NetworkSimplex& upperMap(const UPPER& map) { |
641 | 650 |
delete _pupper; |
642 |
_pupper = new |
|
651 |
_pupper = new FlowArcMap(_graph); |
|
643 | 652 |
for (ArcIt a(_graph); a != INVALID; ++a) { |
644 | 653 |
(*_pupper)[a] = map[a]; |
645 | 654 |
} |
646 | 655 |
return *this; |
647 | 656 |
} |
648 | 657 |
|
649 | 658 |
/// \brief Set the upper bounds (capacities) on the arcs. |
650 | 659 |
/// |
651 | 660 |
/// This function sets the upper bounds (capacities) on the arcs. |
652 | 661 |
/// It is just an alias for \ref upperMap(). |
653 | 662 |
/// |
654 | 663 |
/// \return <tt>(*this)</tt> |
655 | 664 |
template<typename CAP> |
656 | 665 |
NetworkSimplex& capacityMap(const CAP& map) { |
657 | 666 |
return upperMap(map); |
658 | 667 |
} |
659 | 668 |
|
660 | 669 |
/// \brief Set the lower and upper bounds on the arcs. |
661 | 670 |
/// |
662 | 671 |
/// This function sets the lower and upper bounds on the arcs. |
663 | 672 |
/// If neither this function nor \ref lowerMap() is used before |
664 | 673 |
/// calling \ref run(), the lower bounds will be set to zero |
665 | 674 |
/// on all arcs. |
666 | 675 |
/// If none of the functions \ref upperMap(), \ref capacityMap() |
667 | 676 |
/// and \ref boundMaps() is used before calling \ref run(), |
668 | 677 |
/// the upper bounds (capacities) will be set to |
669 |
/// \c std::numeric_limits< |
|
678 |
/// \c std::numeric_limits<Flow>::max() on all arcs. |
|
670 | 679 |
/// |
671 | 680 |
/// \param lower An arc map storing the lower bounds. |
672 | 681 |
/// \param upper An arc map storing the upper bounds. |
673 | 682 |
/// |
674 | 683 |
/// The \c Value type of the maps must be convertible to the |
675 |
/// \c |
|
684 |
/// \c Flow type of the algorithm. |
|
676 | 685 |
/// |
677 | 686 |
/// \note This function is just a shortcut of calling \ref lowerMap() |
678 | 687 |
/// and \ref upperMap() separately. |
679 | 688 |
/// |
680 | 689 |
/// \return <tt>(*this)</tt> |
681 | 690 |
template <typename LOWER, typename UPPER> |
682 | 691 |
NetworkSimplex& boundMaps(const LOWER& lower, const UPPER& upper) { |
683 | 692 |
return lowerMap(lower).upperMap(upper); |
684 | 693 |
} |
685 | 694 |
|
686 | 695 |
/// \brief Set the costs of the arcs. |
687 | 696 |
/// |
688 | 697 |
/// This function sets the costs of the arcs. |
689 | 698 |
/// If it is not used before calling \ref run(), the costs |
690 | 699 |
/// will be set to \c 1 on all arcs. |
691 | 700 |
/// |
692 | 701 |
/// \param map An arc map storing the costs. |
693 |
/// Its \c Value type must be convertible to the \c |
|
702 |
/// Its \c Value type must be convertible to the \c Cost type |
|
694 | 703 |
/// of the algorithm. |
695 | 704 |
/// |
696 | 705 |
/// \return <tt>(*this)</tt> |
697 | 706 |
template<typename COST> |
698 | 707 |
NetworkSimplex& costMap(const COST& map) { |
699 | 708 |
delete _pcost; |
700 |
_pcost = new |
|
709 |
_pcost = new CostArcMap(_graph); |
|
701 | 710 |
for (ArcIt a(_graph); a != INVALID; ++a) { |
702 | 711 |
(*_pcost)[a] = map[a]; |
703 | 712 |
} |
704 | 713 |
return *this; |
705 | 714 |
} |
706 | 715 |
|
707 | 716 |
/// \brief Set the supply values of the nodes. |
708 | 717 |
/// |
709 | 718 |
/// This function sets the supply values of the nodes. |
710 | 719 |
/// If neither this function nor \ref stSupply() is used before |
711 | 720 |
/// calling \ref run(), the supply of each node will be set to zero. |
712 | 721 |
/// (It makes sense only if non-zero lower bounds are given.) |
713 | 722 |
/// |
714 | 723 |
/// \param map A node map storing the supply values. |
715 |
/// Its \c Value type must be convertible to the \c |
|
724 |
/// Its \c Value type must be convertible to the \c Flow type |
|
716 | 725 |
/// of the algorithm. |
717 | 726 |
/// |
718 | 727 |
/// \return <tt>(*this)</tt> |
719 | 728 |
template<typename SUP> |
720 | 729 |
NetworkSimplex& supplyMap(const SUP& map) { |
721 | 730 |
delete _psupply; |
722 | 731 |
_pstsup = false; |
723 |
_psupply = new |
|
732 |
_psupply = new FlowNodeMap(_graph); |
|
724 | 733 |
for (NodeIt n(_graph); n != INVALID; ++n) { |
725 | 734 |
(*_psupply)[n] = map[n]; |
726 | 735 |
} |
727 | 736 |
return *this; |
728 | 737 |
} |
729 | 738 |
|
730 | 739 |
/// \brief Set single source and target nodes and a supply value. |
731 | 740 |
/// |
732 | 741 |
/// This function sets a single source node and a single target node |
733 | 742 |
/// and the required flow value. |
734 | 743 |
/// If neither this function nor \ref supplyMap() is used before |
735 | 744 |
/// calling \ref run(), the supply of each node will be set to zero. |
736 | 745 |
/// (It makes sense only if non-zero lower bounds are given.) |
737 | 746 |
/// |
738 | 747 |
/// \param s The source node. |
739 | 748 |
/// \param t The target node. |
740 | 749 |
/// \param k The required amount of flow from node \c s to node \c t |
741 | 750 |
/// (i.e. the supply of \c s and the demand of \c t). |
742 | 751 |
/// |
743 | 752 |
/// \return <tt>(*this)</tt> |
744 |
NetworkSimplex& stSupply(const Node& s, const Node& t, |
|
753 |
NetworkSimplex& stSupply(const Node& s, const Node& t, Flow k) { |
|
745 | 754 |
delete _psupply; |
746 | 755 |
_psupply = NULL; |
747 | 756 |
_pstsup = true; |
748 | 757 |
_psource = s; |
749 | 758 |
_ptarget = t; |
750 | 759 |
_pstflow = k; |
751 | 760 |
return *this; |
752 | 761 |
} |
753 | 762 |
|
754 | 763 |
/// \brief Set the flow map. |
755 | 764 |
/// |
756 | 765 |
/// This function sets the flow map. |
757 | 766 |
/// If it is not used before calling \ref run(), an instance will |
758 | 767 |
/// be allocated automatically. The destructor deallocates this |
759 | 768 |
/// automatically allocated map, of course. |
760 | 769 |
/// |
761 | 770 |
/// \return <tt>(*this)</tt> |
762 | 771 |
NetworkSimplex& flowMap(FlowMap& map) { |
763 | 772 |
if (_local_flow) { |
764 | 773 |
delete _flow_map; |
765 | 774 |
_local_flow = false; |
766 | 775 |
} |
767 | 776 |
_flow_map = ↦ |
768 | 777 |
return *this; |
769 | 778 |
} |
770 | 779 |
|
771 | 780 |
/// \brief Set the potential map. |
772 | 781 |
/// |
773 | 782 |
/// This function sets the potential map, which is used for storing |
774 | 783 |
/// the dual solution. |
775 | 784 |
/// If it is not used before calling \ref run(), an instance will |
776 | 785 |
/// be allocated automatically. The destructor deallocates this |
777 | 786 |
/// automatically allocated map, of course. |
778 | 787 |
/// |
779 | 788 |
/// \return <tt>(*this)</tt> |
780 | 789 |
NetworkSimplex& potentialMap(PotentialMap& map) { |
781 | 790 |
if (_local_potential) { |
782 | 791 |
delete _potential_map; |
783 | 792 |
_local_potential = false; |
784 | 793 |
} |
785 | 794 |
_potential_map = ↦ |
786 | 795 |
return *this; |
787 | 796 |
} |
788 | 797 |
|
789 | 798 |
/// \name Execution Control |
790 | 799 |
/// The algorithm can be executed using \ref run(). |
791 | 800 |
|
792 | 801 |
/// @{ |
793 | 802 |
|
794 | 803 |
/// \brief Run the algorithm. |
795 | 804 |
/// |
796 | 805 |
/// This function runs the algorithm. |
797 | 806 |
/// The paramters can be specified using \ref lowerMap(), |
798 | 807 |
/// \ref upperMap(), \ref capacityMap(), \ref boundMaps(), |
799 | 808 |
/// \ref costMap(), \ref supplyMap() and \ref stSupply() |
800 | 809 |
/// functions. For example, |
801 | 810 |
/// \code |
802 | 811 |
/// NetworkSimplex<ListDigraph> ns(graph); |
803 | 812 |
/// ns.boundMaps(lower, upper).costMap(cost) |
804 | 813 |
/// .supplyMap(sup).run(); |
805 | 814 |
/// \endcode |
806 | 815 |
/// |
807 | 816 |
/// This function can be called more than once. All the parameters |
808 | 817 |
/// that have been given are kept for the next call, unless |
809 | 818 |
/// \ref reset() is called, thus only the modified parameters |
810 | 819 |
/// have to be set again. See \ref reset() for examples. |
811 | 820 |
/// |
812 | 821 |
/// \param pivot_rule The pivot rule that will be used during the |
813 | 822 |
/// algorithm. For more information see \ref PivotRule. |
814 | 823 |
/// |
815 | 824 |
/// \return \c true if a feasible flow can be found. |
816 | 825 |
bool run(PivotRule pivot_rule = BLOCK_SEARCH) { |
817 | 826 |
return init() && start(pivot_rule); |
818 | 827 |
} |
819 | 828 |
|
820 | 829 |
/// \brief Reset all the parameters that have been given before. |
821 | 830 |
/// |
822 | 831 |
/// This function resets all the paramaters that have been given |
823 | 832 |
/// using \ref lowerMap(), \ref upperMap(), \ref capacityMap(), |
824 | 833 |
/// \ref boundMaps(), \ref costMap(), \ref supplyMap() and |
825 | 834 |
/// \ref stSupply() functions before. |
826 | 835 |
/// |
827 | 836 |
/// It is useful for multiple run() calls. If this function is not |
828 | 837 |
/// used, all the parameters given before are kept for the next |
829 | 838 |
/// \ref run() call. |
830 | 839 |
/// |
831 | 840 |
/// For example, |
832 | 841 |
/// \code |
833 | 842 |
/// NetworkSimplex<ListDigraph> ns(graph); |
834 | 843 |
/// |
835 | 844 |
/// // First run |
836 | 845 |
/// ns.lowerMap(lower).capacityMap(cap).costMap(cost) |
837 | 846 |
/// .supplyMap(sup).run(); |
838 | 847 |
/// |
839 | 848 |
/// // Run again with modified cost map (reset() is not called, |
840 | 849 |
/// // so only the cost map have to be set again) |
841 | 850 |
/// cost[e] += 100; |
842 | 851 |
/// ns.costMap(cost).run(); |
843 | 852 |
/// |
844 | 853 |
/// // Run again from scratch using reset() |
845 | 854 |
/// // (the lower bounds will be set to zero on all arcs) |
846 | 855 |
/// ns.reset(); |
847 | 856 |
/// ns.capacityMap(cap).costMap(cost) |
848 | 857 |
/// .supplyMap(sup).run(); |
849 | 858 |
/// \endcode |
850 | 859 |
/// |
851 | 860 |
/// \return <tt>(*this)</tt> |
852 | 861 |
NetworkSimplex& reset() { |
853 | 862 |
delete _plower; |
854 | 863 |
delete _pupper; |
855 | 864 |
delete _pcost; |
856 | 865 |
delete _psupply; |
857 | 866 |
_plower = NULL; |
858 | 867 |
_pupper = NULL; |
859 | 868 |
_pcost = NULL; |
860 | 869 |
_psupply = NULL; |
861 | 870 |
_pstsup = false; |
862 | 871 |
return *this; |
863 | 872 |
} |
864 | 873 |
|
865 | 874 |
/// @} |
866 | 875 |
|
867 | 876 |
/// \name Query Functions |
868 | 877 |
/// The results of the algorithm can be obtained using these |
869 | 878 |
/// functions.\n |
870 | 879 |
/// The \ref run() function must be called before using them. |
871 | 880 |
|
872 | 881 |
/// @{ |
873 | 882 |
|
874 | 883 |
/// \brief Return the total cost of the found flow. |
875 | 884 |
/// |
876 | 885 |
/// This function returns the total cost of the found flow. |
877 |
/// The complexity of the function is |
|
886 |
/// The complexity of the function is O(e). |
|
878 | 887 |
/// |
879 | 888 |
/// \note The return type of the function can be specified as a |
880 | 889 |
/// template parameter. For example, |
881 | 890 |
/// \code |
882 | 891 |
/// ns.totalCost<double>(); |
883 | 892 |
/// \endcode |
884 |
/// It is useful if the total cost cannot be stored in the \c |
|
893 |
/// It is useful if the total cost cannot be stored in the \c Cost |
|
885 | 894 |
/// type of the algorithm, which is the default return type of the |
886 | 895 |
/// function. |
887 | 896 |
/// |
888 | 897 |
/// \pre \ref run() must be called before using this function. |
889 | 898 |
template <typename Num> |
890 | 899 |
Num totalCost() const { |
891 | 900 |
Num c = 0; |
892 | 901 |
if (_pcost) { |
893 | 902 |
for (ArcIt e(_graph); e != INVALID; ++e) |
894 | 903 |
c += (*_flow_map)[e] * (*_pcost)[e]; |
895 | 904 |
} else { |
896 | 905 |
for (ArcIt e(_graph); e != INVALID; ++e) |
897 | 906 |
c += (*_flow_map)[e]; |
898 | 907 |
} |
899 | 908 |
return c; |
900 | 909 |
} |
901 | 910 |
|
902 | 911 |
#ifndef DOXYGEN |
903 |
Value totalCost() const { |
|
904 |
return totalCost<Value>(); |
|
912 |
Cost totalCost() const { |
|
913 |
return totalCost<Cost>(); |
|
905 | 914 |
} |
906 | 915 |
#endif |
907 | 916 |
|
908 | 917 |
/// \brief Return the flow on the given arc. |
909 | 918 |
/// |
910 | 919 |
/// This function returns the flow on the given arc. |
911 | 920 |
/// |
912 | 921 |
/// \pre \ref run() must be called before using this function. |
913 |
|
|
922 |
Flow flow(const Arc& a) const { |
|
914 | 923 |
return (*_flow_map)[a]; |
915 | 924 |
} |
916 | 925 |
|
917 | 926 |
/// \brief Return a const reference to the flow map. |
918 | 927 |
/// |
919 | 928 |
/// This function returns a const reference to an arc map storing |
920 | 929 |
/// the found flow. |
921 | 930 |
/// |
922 | 931 |
/// \pre \ref run() must be called before using this function. |
923 | 932 |
const FlowMap& flowMap() const { |
924 | 933 |
return *_flow_map; |
925 | 934 |
} |
926 | 935 |
|
927 | 936 |
/// \brief Return the potential (dual value) of the given node. |
928 | 937 |
/// |
929 | 938 |
/// This function returns the potential (dual value) of the |
930 | 939 |
/// given node. |
931 | 940 |
/// |
932 | 941 |
/// \pre \ref run() must be called before using this function. |
933 |
|
|
942 |
Cost potential(const Node& n) const { |
|
934 | 943 |
return (*_potential_map)[n]; |
935 | 944 |
} |
936 | 945 |
|
937 | 946 |
/// \brief Return a const reference to the potential map |
938 | 947 |
/// (the dual solution). |
939 | 948 |
/// |
940 | 949 |
/// This function returns a const reference to a node map storing |
941 | 950 |
/// the found potentials, which form the dual solution of the |
942 | 951 |
/// \ref min_cost_flow "minimum cost flow" problem. |
943 | 952 |
/// |
944 | 953 |
/// \pre \ref run() must be called before using this function. |
945 | 954 |
const PotentialMap& potentialMap() const { |
946 | 955 |
return *_potential_map; |
947 | 956 |
} |
948 | 957 |
|
949 | 958 |
/// @} |
950 | 959 |
|
951 | 960 |
private: |
952 | 961 |
|
953 | 962 |
// Initialize internal data structures |
954 | 963 |
bool init() { |
955 | 964 |
// Initialize result maps |
956 | 965 |
if (!_flow_map) { |
957 | 966 |
_flow_map = new FlowMap(_graph); |
958 | 967 |
_local_flow = true; |
959 | 968 |
} |
960 | 969 |
if (!_potential_map) { |
961 | 970 |
_potential_map = new PotentialMap(_graph); |
962 | 971 |
_local_potential = true; |
963 | 972 |
} |
964 | 973 |
|
965 | 974 |
// Initialize vectors |
966 | 975 |
_node_num = countNodes(_graph); |
967 | 976 |
_arc_num = countArcs(_graph); |
968 | 977 |
int all_node_num = _node_num + 1; |
969 | 978 |
int all_arc_num = _arc_num + _node_num; |
970 | 979 |
if (_node_num == 0) return false; |
971 | 980 |
|
972 | 981 |
_arc_ref.resize(_arc_num); |
973 | 982 |
_source.resize(all_arc_num); |
974 | 983 |
_target.resize(all_arc_num); |
975 | 984 |
|
976 | 985 |
_cap.resize(all_arc_num); |
977 | 986 |
_cost.resize(all_arc_num); |
978 | 987 |
_supply.resize(all_node_num); |
979 | 988 |
_flow.resize(all_arc_num); |
980 | 989 |
_pi.resize(all_node_num); |
981 | 990 |
|
982 | 991 |
_parent.resize(all_node_num); |
983 | 992 |
_pred.resize(all_node_num); |
984 | 993 |
_forward.resize(all_node_num); |
985 | 994 |
_thread.resize(all_node_num); |
986 | 995 |
_rev_thread.resize(all_node_num); |
987 | 996 |
_succ_num.resize(all_node_num); |
988 | 997 |
_last_succ.resize(all_node_num); |
989 | 998 |
_state.resize(all_arc_num); |
990 | 999 |
|
991 | 1000 |
// Initialize node related data |
992 | 1001 |
bool valid_supply = true; |
993 | 1002 |
if (!_pstsup && !_psupply) { |
994 | 1003 |
_pstsup = true; |
995 | 1004 |
_psource = _ptarget = NodeIt(_graph); |
996 | 1005 |
_pstflow = 0; |
997 | 1006 |
} |
998 | 1007 |
if (_psupply) { |
999 |
|
|
1008 |
Flow sum = 0; |
|
1000 | 1009 |
int i = 0; |
1001 | 1010 |
for (NodeIt n(_graph); n != INVALID; ++n, ++i) { |
1002 | 1011 |
_node_id[n] = i; |
1003 | 1012 |
_supply[i] = (*_psupply)[n]; |
1004 | 1013 |
sum += _supply[i]; |
1005 | 1014 |
} |
1006 | 1015 |
valid_supply = (sum == 0); |
1007 | 1016 |
} else { |
1008 | 1017 |
int i = 0; |
1009 | 1018 |
for (NodeIt n(_graph); n != INVALID; ++n, ++i) { |
1010 | 1019 |
_node_id[n] = i; |
1011 | 1020 |
_supply[i] = 0; |
1012 | 1021 |
} |
1013 | 1022 |
_supply[_node_id[_psource]] = _pstflow; |
1014 | 1023 |
_supply[_node_id[_ptarget]] = -_pstflow; |
1015 | 1024 |
} |
1016 | 1025 |
if (!valid_supply) return false; |
1017 | 1026 |
|
1018 | 1027 |
// Set data for the artificial root node |
1019 | 1028 |
_root = _node_num; |
1020 | 1029 |
_parent[_root] = -1; |
1021 | 1030 |
_pred[_root] = -1; |
1022 | 1031 |
_thread[_root] = 0; |
1023 | 1032 |
_rev_thread[0] = _root; |
1024 | 1033 |
_succ_num[_root] = all_node_num; |
1025 | 1034 |
_last_succ[_root] = _root - 1; |
1026 | 1035 |
_supply[_root] = 0; |
1027 | 1036 |
_pi[_root] = 0; |
1028 | 1037 |
|
1029 | 1038 |
// Store the arcs in a mixed order |
1030 | 1039 |
int k = std::max(int(sqrt(_arc_num)), 10); |
1031 | 1040 |
int i = 0; |
1032 | 1041 |
for (ArcIt e(_graph); e != INVALID; ++e) { |
1033 | 1042 |
_arc_ref[i] = e; |
1034 | 1043 |
if ((i += k) >= _arc_num) i = (i % k) + 1; |
1035 | 1044 |
} |
1036 | 1045 |
|
1037 | 1046 |
// Initialize arc maps |
1047 |
Flow max_cap = std::numeric_limits<Flow>::max(); |
|
1048 |
Cost max_cost = std::numeric_limits<Cost>::max() / 4; |
|
1038 | 1049 |
if (_pupper && _pcost) { |
1039 | 1050 |
for (int i = 0; i != _arc_num; ++i) { |
1040 | 1051 |
Arc e = _arc_ref[i]; |
1041 | 1052 |
_source[i] = _node_id[_graph.source(e)]; |
1042 | 1053 |
_target[i] = _node_id[_graph.target(e)]; |
1043 | 1054 |
_cap[i] = (*_pupper)[e]; |
1044 | 1055 |
_cost[i] = (*_pcost)[e]; |
1045 | 1056 |
_flow[i] = 0; |
1046 | 1057 |
_state[i] = STATE_LOWER; |
1047 | 1058 |
} |
1048 | 1059 |
} else { |
1049 | 1060 |
for (int i = 0; i != _arc_num; ++i) { |
1050 | 1061 |
Arc e = _arc_ref[i]; |
1051 | 1062 |
_source[i] = _node_id[_graph.source(e)]; |
1052 | 1063 |
_target[i] = _node_id[_graph.target(e)]; |
1053 | 1064 |
_flow[i] = 0; |
1054 | 1065 |
_state[i] = STATE_LOWER; |
1055 | 1066 |
} |
1056 | 1067 |
if (_pupper) { |
1057 | 1068 |
for (int i = 0; i != _arc_num; ++i) |
1058 | 1069 |
_cap[i] = (*_pupper)[_arc_ref[i]]; |
1059 | 1070 |
} else { |
1060 |
Value val = std::numeric_limits<Value>::max(); |
|
1061 | 1071 |
for (int i = 0; i != _arc_num; ++i) |
1062 |
_cap[i] = |
|
1072 |
_cap[i] = max_cap; |
|
1063 | 1073 |
} |
1064 | 1074 |
if (_pcost) { |
1065 | 1075 |
for (int i = 0; i != _arc_num; ++i) |
1066 | 1076 |
_cost[i] = (*_pcost)[_arc_ref[i]]; |
1067 | 1077 |
} else { |
1068 | 1078 |
for (int i = 0; i != _arc_num; ++i) |
1069 | 1079 |
_cost[i] = 1; |
1070 | 1080 |
} |
1071 | 1081 |
} |
1072 | 1082 |
|
1073 | 1083 |
// Remove non-zero lower bounds |
1074 | 1084 |
if (_plower) { |
1075 | 1085 |
for (int i = 0; i != _arc_num; ++i) { |
1076 |
|
|
1086 |
Flow c = (*_plower)[_arc_ref[i]]; |
|
1077 | 1087 |
if (c != 0) { |
1078 | 1088 |
_cap[i] -= c; |
1079 | 1089 |
_supply[_source[i]] -= c; |
1080 | 1090 |
_supply[_target[i]] += c; |
1081 | 1091 |
} |
1082 | 1092 |
} |
1083 | 1093 |
} |
1084 | 1094 |
|
1085 | 1095 |
// Add artificial arcs and initialize the spanning tree data structure |
1086 |
Value max_cap = std::numeric_limits<Value>::max(); |
|
1087 |
Value max_cost = std::numeric_limits<Value>::max() / 4; |
|
1088 | 1096 |
for (int u = 0, e = _arc_num; u != _node_num; ++u, ++e) { |
1089 | 1097 |
_thread[u] = u + 1; |
1090 | 1098 |
_rev_thread[u + 1] = u; |
1091 | 1099 |
_succ_num[u] = 1; |
1092 | 1100 |
_last_succ[u] = u; |
1093 | 1101 |
_parent[u] = _root; |
1094 | 1102 |
_pred[u] = e; |
1095 | 1103 |
_cost[e] = max_cost; |
1096 | 1104 |
_cap[e] = max_cap; |
1097 | 1105 |
_state[e] = STATE_TREE; |
1098 | 1106 |
if (_supply[u] >= 0) { |
1099 | 1107 |
_flow[e] = _supply[u]; |
1100 | 1108 |
_forward[u] = true; |
1101 | 1109 |
_pi[u] = -max_cost; |
1102 | 1110 |
} else { |
1103 | 1111 |
_flow[e] = -_supply[u]; |
1104 | 1112 |
_forward[u] = false; |
1105 | 1113 |
_pi[u] = max_cost; |
1106 | 1114 |
} |
1107 | 1115 |
} |
1108 | 1116 |
|
1109 | 1117 |
return true; |
1110 | 1118 |
} |
1111 | 1119 |
|
1112 | 1120 |
// Find the join node |
1113 | 1121 |
void findJoinNode() { |
1114 | 1122 |
int u = _source[in_arc]; |
1115 | 1123 |
int v = _target[in_arc]; |
1116 | 1124 |
while (u != v) { |
1117 | 1125 |
if (_succ_num[u] < _succ_num[v]) { |
1118 | 1126 |
u = _parent[u]; |
1119 | 1127 |
} else { |
1120 | 1128 |
v = _parent[v]; |
1121 | 1129 |
} |
1122 | 1130 |
} |
1123 | 1131 |
join = u; |
1124 | 1132 |
} |
1125 | 1133 |
|
1126 | 1134 |
// Find the leaving arc of the cycle and returns true if the |
1127 | 1135 |
// leaving arc is not the same as the entering arc |
1128 | 1136 |
bool findLeavingArc() { |
1129 | 1137 |
// Initialize first and second nodes according to the direction |
1130 | 1138 |
// of the cycle |
1131 | 1139 |
if (_state[in_arc] == STATE_LOWER) { |
1132 | 1140 |
first = _source[in_arc]; |
1133 | 1141 |
second = _target[in_arc]; |
1134 | 1142 |
} else { |
1135 | 1143 |
first = _target[in_arc]; |
1136 | 1144 |
second = _source[in_arc]; |
1137 | 1145 |
} |
1138 | 1146 |
delta = _cap[in_arc]; |
1139 | 1147 |
int result = 0; |
1140 |
|
|
1148 |
Flow d; |
|
1141 | 1149 |
int e; |
1142 | 1150 |
|
1143 | 1151 |
// Search the cycle along the path form the first node to the root |
1144 | 1152 |
for (int u = first; u != join; u = _parent[u]) { |
1145 | 1153 |
e = _pred[u]; |
1146 | 1154 |
d = _forward[u] ? _flow[e] : _cap[e] - _flow[e]; |
1147 | 1155 |
if (d < delta) { |
1148 | 1156 |
delta = d; |
1149 | 1157 |
u_out = u; |
1150 | 1158 |
result = 1; |
1151 | 1159 |
} |
1152 | 1160 |
} |
1153 | 1161 |
// Search the cycle along the path form the second node to the root |
1154 | 1162 |
for (int u = second; u != join; u = _parent[u]) { |
1155 | 1163 |
e = _pred[u]; |
1156 | 1164 |
d = _forward[u] ? _cap[e] - _flow[e] : _flow[e]; |
1157 | 1165 |
if (d <= delta) { |
1158 | 1166 |
delta = d; |
1159 | 1167 |
u_out = u; |
1160 | 1168 |
result = 2; |
1161 | 1169 |
} |
1162 | 1170 |
} |
1163 | 1171 |
|
1164 | 1172 |
if (result == 1) { |
1165 | 1173 |
u_in = first; |
1166 | 1174 |
v_in = second; |
1167 | 1175 |
} else { |
1168 | 1176 |
u_in = second; |
1169 | 1177 |
v_in = first; |
1170 | 1178 |
} |
1171 | 1179 |
return result != 0; |
1172 | 1180 |
} |
1173 | 1181 |
|
1174 | 1182 |
// Change _flow and _state vectors |
1175 | 1183 |
void changeFlow(bool change) { |
1176 | 1184 |
// Augment along the cycle |
1177 | 1185 |
if (delta > 0) { |
1178 |
|
|
1186 |
Flow val = _state[in_arc] * delta; |
|
1179 | 1187 |
_flow[in_arc] += val; |
1180 | 1188 |
for (int u = _source[in_arc]; u != join; u = _parent[u]) { |
1181 | 1189 |
_flow[_pred[u]] += _forward[u] ? -val : val; |
1182 | 1190 |
} |
1183 | 1191 |
for (int u = _target[in_arc]; u != join; u = _parent[u]) { |
1184 | 1192 |
_flow[_pred[u]] += _forward[u] ? val : -val; |
1185 | 1193 |
} |
1186 | 1194 |
} |
1187 | 1195 |
// Update the state of the entering and leaving arcs |
1188 | 1196 |
if (change) { |
1189 | 1197 |
_state[in_arc] = STATE_TREE; |
1190 | 1198 |
_state[_pred[u_out]] = |
1191 | 1199 |
(_flow[_pred[u_out]] == 0) ? STATE_LOWER : STATE_UPPER; |
1192 | 1200 |
} else { |
1193 | 1201 |
_state[in_arc] = -_state[in_arc]; |
1194 | 1202 |
} |
1195 | 1203 |
} |
1196 | 1204 |
|
1197 | 1205 |
// Update the tree structure |
1198 | 1206 |
void updateTreeStructure() { |
1199 | 1207 |
int u, w; |
1200 | 1208 |
int old_rev_thread = _rev_thread[u_out]; |
1201 | 1209 |
int old_succ_num = _succ_num[u_out]; |
1202 | 1210 |
int old_last_succ = _last_succ[u_out]; |
1203 | 1211 |
v_out = _parent[u_out]; |
1204 | 1212 |
|
1205 | 1213 |
u = _last_succ[u_in]; // the last successor of u_in |
1206 | 1214 |
right = _thread[u]; // the node after it |
1207 | 1215 |
|
1208 | 1216 |
// Handle the case when old_rev_thread equals to v_in |
1209 | 1217 |
// (it also means that join and v_out coincide) |
1210 | 1218 |
if (old_rev_thread == v_in) { |
1211 | 1219 |
last = _thread[_last_succ[u_out]]; |
1212 | 1220 |
} else { |
1213 | 1221 |
last = _thread[v_in]; |
1214 | 1222 |
} |
1215 | 1223 |
|
1216 | 1224 |
// Update _thread and _parent along the stem nodes (i.e. the nodes |
1217 | 1225 |
// between u_in and u_out, whose parent have to be changed) |
1218 | 1226 |
_thread[v_in] = stem = u_in; |
1219 | 1227 |
_dirty_revs.clear(); |
1220 | 1228 |
_dirty_revs.push_back(v_in); |
1221 | 1229 |
par_stem = v_in; |
1222 | 1230 |
while (stem != u_out) { |
1223 | 1231 |
// Insert the next stem node into the thread list |
1224 | 1232 |
new_stem = _parent[stem]; |
1225 | 1233 |
_thread[u] = new_stem; |
1226 | 1234 |
_dirty_revs.push_back(u); |
1227 | 1235 |
|
1228 | 1236 |
// Remove the subtree of stem from the thread list |
1229 | 1237 |
w = _rev_thread[stem]; |
1230 | 1238 |
_thread[w] = right; |
1231 | 1239 |
_rev_thread[right] = w; |
1232 | 1240 |
|
1233 | 1241 |
// Change the parent node and shift stem nodes |
1234 | 1242 |
_parent[stem] = par_stem; |
1235 | 1243 |
par_stem = stem; |
1236 | 1244 |
stem = new_stem; |
1237 | 1245 |
|
1238 | 1246 |
// Update u and right |
1239 | 1247 |
u = _last_succ[stem] == _last_succ[par_stem] ? |
1240 | 1248 |
_rev_thread[par_stem] : _last_succ[stem]; |
1241 | 1249 |
right = _thread[u]; |
1242 | 1250 |
} |
1243 | 1251 |
_parent[u_out] = par_stem; |
1244 | 1252 |
_thread[u] = last; |
1245 | 1253 |
_rev_thread[last] = u; |
1246 | 1254 |
_last_succ[u_out] = u; |
1247 | 1255 |
|
1248 | 1256 |
// Remove the subtree of u_out from the thread list except for |
1249 | 1257 |
// the case when old_rev_thread equals to v_in |
1250 | 1258 |
// (it also means that join and v_out coincide) |
1251 | 1259 |
if (old_rev_thread != v_in) { |
1252 | 1260 |
_thread[old_rev_thread] = right; |
1253 | 1261 |
_rev_thread[right] = old_rev_thread; |
1254 | 1262 |
} |
1255 | 1263 |
|
1256 | 1264 |
// Update _rev_thread using the new _thread values |
1257 | 1265 |
for (int i = 0; i < int(_dirty_revs.size()); ++i) { |
1258 | 1266 |
u = _dirty_revs[i]; |
1259 | 1267 |
_rev_thread[_thread[u]] = u; |
1260 | 1268 |
} |
1261 | 1269 |
|
1262 | 1270 |
// Update _pred, _forward, _last_succ and _succ_num for the |
1263 | 1271 |
// stem nodes from u_out to u_in |
1264 | 1272 |
int tmp_sc = 0, tmp_ls = _last_succ[u_out]; |
1265 | 1273 |
u = u_out; |
1266 | 1274 |
while (u != u_in) { |
1267 | 1275 |
w = _parent[u]; |
1268 | 1276 |
_pred[u] = _pred[w]; |
1269 | 1277 |
_forward[u] = !_forward[w]; |
1270 | 1278 |
tmp_sc += _succ_num[u] - _succ_num[w]; |
1271 | 1279 |
_succ_num[u] = tmp_sc; |
1272 | 1280 |
_last_succ[w] = tmp_ls; |
1273 | 1281 |
u = w; |
1274 | 1282 |
} |
1275 | 1283 |
_pred[u_in] = in_arc; |
1276 | 1284 |
_forward[u_in] = (u_in == _source[in_arc]); |
1277 | 1285 |
_succ_num[u_in] = old_succ_num; |
1278 | 1286 |
|
1279 | 1287 |
// Set limits for updating _last_succ form v_in and v_out |
1280 | 1288 |
// towards the root |
1281 | 1289 |
int up_limit_in = -1; |
1282 | 1290 |
int up_limit_out = -1; |
1283 | 1291 |
if (_last_succ[join] == v_in) { |
1284 | 1292 |
up_limit_out = join; |
1285 | 1293 |
} else { |
1286 | 1294 |
up_limit_in = join; |
1287 | 1295 |
} |
1288 | 1296 |
|
1289 | 1297 |
// Update _last_succ from v_in towards the root |
1290 | 1298 |
for (u = v_in; u != up_limit_in && _last_succ[u] == v_in; |
1291 | 1299 |
u = _parent[u]) { |
1292 | 1300 |
_last_succ[u] = _last_succ[u_out]; |
1293 | 1301 |
} |
1294 | 1302 |
// Update _last_succ from v_out towards the root |
1295 | 1303 |
if (join != old_rev_thread && v_in != old_rev_thread) { |
1296 | 1304 |
for (u = v_out; u != up_limit_out && _last_succ[u] == old_last_succ; |
1297 | 1305 |
u = _parent[u]) { |
1298 | 1306 |
_last_succ[u] = old_rev_thread; |
1299 | 1307 |
} |
1300 | 1308 |
} else { |
1301 | 1309 |
for (u = v_out; u != up_limit_out && _last_succ[u] == old_last_succ; |
1302 | 1310 |
u = _parent[u]) { |
1303 | 1311 |
_last_succ[u] = _last_succ[u_out]; |
1304 | 1312 |
} |
1305 | 1313 |
} |
1306 | 1314 |
|
1307 | 1315 |
// Update _succ_num from v_in to join |
1308 | 1316 |
for (u = v_in; u != join; u = _parent[u]) { |
1309 | 1317 |
_succ_num[u] += old_succ_num; |
1310 | 1318 |
} |
1311 | 1319 |
// Update _succ_num from v_out to join |
1312 | 1320 |
for (u = v_out; u != join; u = _parent[u]) { |
1313 | 1321 |
_succ_num[u] -= old_succ_num; |
1314 | 1322 |
} |
1315 | 1323 |
} |
1316 | 1324 |
|
1317 | 1325 |
// Update potentials |
1318 | 1326 |
void updatePotential() { |
1319 |
|
|
1327 |
Cost sigma = _forward[u_in] ? |
|
1320 | 1328 |
_pi[v_in] - _pi[u_in] - _cost[_pred[u_in]] : |
1321 | 1329 |
_pi[v_in] - _pi[u_in] + _cost[_pred[u_in]]; |
1322 | 1330 |
if (_succ_num[u_in] > _node_num / 2) { |
1323 | 1331 |
// Update in the upper subtree (which contains the root) |
1324 | 1332 |
int before = _rev_thread[u_in]; |
1325 | 1333 |
int after = _thread[_last_succ[u_in]]; |
1326 | 1334 |
_thread[before] = after; |
1327 | 1335 |
_pi[_root] -= sigma; |
1328 | 1336 |
for (int u = _thread[_root]; u != _root; u = _thread[u]) { |
1329 | 1337 |
_pi[u] -= sigma; |
1330 | 1338 |
} |
1331 | 1339 |
_thread[before] = u_in; |
1332 | 1340 |
} else { |
1333 | 1341 |
// Update in the lower subtree (which has been moved) |
1334 | 1342 |
int end = _thread[_last_succ[u_in]]; |
1335 | 1343 |
for (int u = u_in; u != end; u = _thread[u]) { |
1336 | 1344 |
_pi[u] += sigma; |
1337 | 1345 |
} |
1338 | 1346 |
} |
1339 | 1347 |
} |
1340 | 1348 |
|
1341 | 1349 |
// Execute the algorithm |
1342 | 1350 |
bool start(PivotRule pivot_rule) { |
1343 | 1351 |
// Select the pivot rule implementation |
1344 | 1352 |
switch (pivot_rule) { |
1345 | 1353 |
case FIRST_ELIGIBLE: |
1346 | 1354 |
return start<FirstEligiblePivotRule>(); |
1347 | 1355 |
case BEST_ELIGIBLE: |
1348 | 1356 |
return start<BestEligiblePivotRule>(); |
1349 | 1357 |
case BLOCK_SEARCH: |
1350 | 1358 |
return start<BlockSearchPivotRule>(); |
1351 | 1359 |
case CANDIDATE_LIST: |
1352 | 1360 |
return start<CandidateListPivotRule>(); |
1353 | 1361 |
case ALTERING_LIST: |
1354 | 1362 |
return start<AlteringListPivotRule>(); |
1355 | 1363 |
} |
1356 | 1364 |
return false; |
1357 | 1365 |
} |
1358 | 1366 |
|
1359 | 1367 |
template <typename PivotRuleImpl> |
1360 | 1368 |
bool start() { |
1361 | 1369 |
PivotRuleImpl pivot(*this); |
1362 | 1370 |
|
1363 | 1371 |
// Execute the Network Simplex algorithm |
1364 | 1372 |
while (pivot.findEnteringArc()) { |
1365 | 1373 |
findJoinNode(); |
1366 | 1374 |
bool change = findLeavingArc(); |
1367 | 1375 |
changeFlow(change); |
1368 | 1376 |
if (change) { |
1369 | 1377 |
updateTreeStructure(); |
1370 | 1378 |
updatePotential(); |
1371 | 1379 |
} |
1372 | 1380 |
} |
1373 | 1381 |
|
1374 | 1382 |
// Check if the flow amount equals zero on all the artificial arcs |
1375 | 1383 |
for (int e = _arc_num; e != _arc_num + _node_num; ++e) { |
1376 | 1384 |
if (_flow[e] > 0) return false; |
1377 | 1385 |
} |
1378 | 1386 |
|
1379 | 1387 |
// Copy flow values to _flow_map |
1380 | 1388 |
if (_plower) { |
1381 | 1389 |
for (int i = 0; i != _arc_num; ++i) { |
1382 | 1390 |
Arc e = _arc_ref[i]; |
1383 | 1391 |
_flow_map->set(e, (*_plower)[e] + _flow[i]); |
1384 | 1392 |
} |
1385 | 1393 |
} else { |
1386 | 1394 |
for (int i = 0; i != _arc_num; ++i) { |
1387 | 1395 |
_flow_map->set(_arc_ref[i], _flow[i]); |
1388 | 1396 |
} |
1389 | 1397 |
} |
1390 | 1398 |
// Copy potential values to _potential_map |
1391 | 1399 |
for (NodeIt n(_graph); n != INVALID; ++n) { |
1392 | 1400 |
_potential_map->set(n, _pi[_node_id[n]]); |
1393 | 1401 |
} |
1394 | 1402 |
|
1395 | 1403 |
return true; |
1396 | 1404 |
} |
1397 | 1405 |
|
1398 | 1406 |
}; //class NetworkSimplex |
1399 | 1407 |
|
1400 | 1408 |
///@} |
1401 | 1409 |
|
1402 | 1410 |
} //namespace lemon |
1403 | 1411 |
|
1404 | 1412 |
#endif //LEMON_NETWORK_SIMPLEX_H |
1 | 1 |
/* -*- mode: C++; indent-tabs-mode: nil; -*- |
2 | 2 |
* |
3 | 3 |
* This file is a part of LEMON, a generic C++ optimization library. |
4 | 4 |
* |
5 | 5 |
* Copyright (C) 2003-2009 |
6 | 6 |
* Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport |
7 | 7 |
* (Egervary Research Group on Combinatorial Optimization, EGRES). |
8 | 8 |
* |
9 | 9 |
* Permission to use, modify and distribute this software is granted |
10 | 10 |
* provided that this copyright notice appears in all copies. For |
11 | 11 |
* precise terms see the accompanying LICENSE file. |
12 | 12 |
* |
13 | 13 |
* This software is provided "AS IS" with no warranty of any kind, |
14 | 14 |
* express or implied, and with no claim as to its suitability for any |
15 | 15 |
* purpose. |
16 | 16 |
* |
17 | 17 |
*/ |
18 | 18 |
|
19 | 19 |
#include <iostream> |
20 | 20 |
#include <fstream> |
21 | 21 |
|
22 | 22 |
#include <lemon/list_graph.h> |
23 | 23 |
#include <lemon/lgf_reader.h> |
24 | 24 |
|
25 | 25 |
#include <lemon/network_simplex.h> |
26 | 26 |
|
27 | 27 |
#include <lemon/concepts/digraph.h> |
28 | 28 |
#include <lemon/concept_check.h> |
29 | 29 |
|
30 | 30 |
#include "test_tools.h" |
31 | 31 |
|
32 | 32 |
using namespace lemon; |
33 | 33 |
|
34 | 34 |
char test_lgf[] = |
35 | 35 |
"@nodes\n" |
36 | 36 |
"label sup1 sup2 sup3\n" |
37 | 37 |
" 1 20 27 0\n" |
38 | 38 |
" 2 -4 0 0\n" |
39 | 39 |
" 3 0 0 0\n" |
40 | 40 |
" 4 0 0 0\n" |
41 | 41 |
" 5 9 0 0\n" |
42 | 42 |
" 6 -6 0 0\n" |
43 | 43 |
" 7 0 0 0\n" |
44 | 44 |
" 8 0 0 0\n" |
45 | 45 |
" 9 3 0 0\n" |
46 | 46 |
" 10 -2 0 0\n" |
47 | 47 |
" 11 0 0 0\n" |
48 | 48 |
" 12 -20 -27 0\n" |
49 | 49 |
"\n" |
50 | 50 |
"@arcs\n" |
51 | 51 |
" cost cap low1 low2\n" |
52 | 52 |
" 1 2 70 11 0 8\n" |
53 | 53 |
" 1 3 150 3 0 1\n" |
54 | 54 |
" 1 4 80 15 0 2\n" |
55 | 55 |
" 2 8 80 12 0 0\n" |
56 | 56 |
" 3 5 140 5 0 3\n" |
57 | 57 |
" 4 6 60 10 0 1\n" |
58 | 58 |
" 4 7 80 2 0 0\n" |
59 | 59 |
" 4 8 110 3 0 0\n" |
60 | 60 |
" 5 7 60 14 0 0\n" |
61 | 61 |
" 5 11 120 12 0 0\n" |
62 | 62 |
" 6 3 0 3 0 0\n" |
63 | 63 |
" 6 9 140 4 0 0\n" |
64 | 64 |
" 6 10 90 8 0 0\n" |
65 | 65 |
" 7 1 30 5 0 0\n" |
66 | 66 |
" 8 12 60 16 0 4\n" |
67 | 67 |
" 9 12 50 6 0 0\n" |
68 | 68 |
"10 12 70 13 0 5\n" |
69 | 69 |
"10 2 100 7 0 0\n" |
70 | 70 |
"10 7 60 10 0 0\n" |
71 | 71 |
"11 10 20 14 0 6\n" |
72 | 72 |
"12 11 30 10 0 0\n" |
73 | 73 |
"\n" |
74 | 74 |
"@attributes\n" |
75 | 75 |
"source 1\n" |
76 | 76 |
"target 12\n"; |
77 | 77 |
|
78 | 78 |
|
79 | 79 |
// Check the interface of an MCF algorithm |
80 |
template <typename GR, typename |
|
80 |
template <typename GR, typename Flow, typename Cost> |
|
81 | 81 |
class McfClassConcept |
82 | 82 |
{ |
83 | 83 |
public: |
84 | 84 |
|
85 | 85 |
template <typename MCF> |
86 | 86 |
struct Constraints { |
87 | 87 |
void constraints() { |
88 | 88 |
checkConcept<concepts::Digraph, GR>(); |
89 | 89 |
|
90 | 90 |
MCF mcf(g); |
91 | 91 |
|
92 | 92 |
b = mcf.reset() |
93 | 93 |
.lowerMap(lower) |
94 | 94 |
.upperMap(upper) |
95 | 95 |
.capacityMap(upper) |
96 | 96 |
.boundMaps(lower, upper) |
97 | 97 |
.costMap(cost) |
98 | 98 |
.supplyMap(sup) |
99 | 99 |
.stSupply(n, n, k) |
100 | 100 |
.run(); |
101 | 101 |
|
102 | 102 |
const typename MCF::FlowMap &fm = mcf.flowMap(); |
103 | 103 |
const typename MCF::PotentialMap &pm = mcf.potentialMap(); |
104 | 104 |
|
105 | 105 |
v = mcf.totalCost(); |
106 | 106 |
double x = mcf.template totalCost<double>(); |
107 | 107 |
v = mcf.flow(a); |
108 | 108 |
v = mcf.potential(n); |
109 | 109 |
mcf.flowMap(flow); |
110 | 110 |
mcf.potentialMap(pot); |
111 | 111 |
|
112 | 112 |
ignore_unused_variable_warning(fm); |
113 | 113 |
ignore_unused_variable_warning(pm); |
114 | 114 |
ignore_unused_variable_warning(x); |
115 | 115 |
} |
116 | 116 |
|
117 | 117 |
typedef typename GR::Node Node; |
118 | 118 |
typedef typename GR::Arc Arc; |
119 |
typedef concepts::ReadMap<Node, Value> NM; |
|
120 |
typedef concepts::ReadMap<Arc, Value> AM; |
|
119 |
typedef concepts::ReadMap<Node, Flow> NM; |
|
120 |
typedef concepts::ReadMap<Arc, Flow> FAM; |
|
121 |
typedef concepts::ReadMap<Arc, Cost> CAM; |
|
121 | 122 |
|
122 | 123 |
const GR &g; |
123 |
const AM &lower; |
|
124 |
const AM &upper; |
|
125 |
const |
|
124 |
const FAM &lower; |
|
125 |
const FAM &upper; |
|
126 |
const CAM &cost; |
|
126 | 127 |
const NM ⊃ |
127 | 128 |
const Node &n; |
128 | 129 |
const Arc &a; |
129 |
const Value &k; |
|
130 |
Value v; |
|
130 |
const Flow &k; |
|
131 |
Flow v; |
|
131 | 132 |
bool b; |
132 | 133 |
|
133 | 134 |
typename MCF::FlowMap &flow; |
134 | 135 |
typename MCF::PotentialMap &pot; |
135 | 136 |
}; |
136 | 137 |
|
137 | 138 |
}; |
138 | 139 |
|
139 | 140 |
|
140 | 141 |
// Check the feasibility of the given flow (primal soluiton) |
141 | 142 |
template < typename GR, typename LM, typename UM, |
142 | 143 |
typename SM, typename FM > |
143 | 144 |
bool checkFlow( const GR& gr, const LM& lower, const UM& upper, |
144 | 145 |
const SM& supply, const FM& flow ) |
145 | 146 |
{ |
146 | 147 |
TEMPLATE_DIGRAPH_TYPEDEFS(GR); |
147 | 148 |
|
148 | 149 |
for (ArcIt e(gr); e != INVALID; ++e) { |
149 | 150 |
if (flow[e] < lower[e] || flow[e] > upper[e]) return false; |
150 | 151 |
} |
151 | 152 |
|
152 | 153 |
for (NodeIt n(gr); n != INVALID; ++n) { |
153 | 154 |
typename SM::Value sum = 0; |
154 | 155 |
for (OutArcIt e(gr, n); e != INVALID; ++e) |
155 | 156 |
sum += flow[e]; |
156 | 157 |
for (InArcIt e(gr, n); e != INVALID; ++e) |
157 | 158 |
sum -= flow[e]; |
158 | 159 |
if (sum != supply[n]) return false; |
159 | 160 |
} |
160 | 161 |
|
161 | 162 |
return true; |
162 | 163 |
} |
163 | 164 |
|
164 | 165 |
// Check the feasibility of the given potentials (dual soluiton) |
165 | 166 |
// using the "Complementary Slackness" optimality condition |
166 | 167 |
template < typename GR, typename LM, typename UM, |
167 | 168 |
typename CM, typename FM, typename PM > |
168 | 169 |
bool checkPotential( const GR& gr, const LM& lower, const UM& upper, |
169 | 170 |
const CM& cost, const FM& flow, const PM& pi ) |
170 | 171 |
{ |
171 | 172 |
TEMPLATE_DIGRAPH_TYPEDEFS(GR); |
172 | 173 |
|
173 | 174 |
bool opt = true; |
174 | 175 |
for (ArcIt e(gr); opt && e != INVALID; ++e) { |
175 | 176 |
typename CM::Value red_cost = |
176 | 177 |
cost[e] + pi[gr.source(e)] - pi[gr.target(e)]; |
177 | 178 |
opt = red_cost == 0 || |
178 | 179 |
(red_cost > 0 && flow[e] == lower[e]) || |
179 | 180 |
(red_cost < 0 && flow[e] == upper[e]); |
180 | 181 |
} |
181 | 182 |
return opt; |
182 | 183 |
} |
183 | 184 |
|
184 | 185 |
// Run a minimum cost flow algorithm and check the results |
185 | 186 |
template < typename MCF, typename GR, |
186 | 187 |
typename LM, typename UM, |
187 | 188 |
typename CM, typename SM > |
188 | 189 |
void checkMcf( const MCF& mcf, bool mcf_result, |
189 | 190 |
const GR& gr, const LM& lower, const UM& upper, |
190 | 191 |
const CM& cost, const SM& supply, |
191 | 192 |
bool result, typename CM::Value total, |
192 | 193 |
const std::string &test_id = "" ) |
193 | 194 |
{ |
194 | 195 |
check(mcf_result == result, "Wrong result " + test_id); |
195 | 196 |
if (result) { |
196 | 197 |
check(checkFlow(gr, lower, upper, supply, mcf.flowMap()), |
197 | 198 |
"The flow is not feasible " + test_id); |
198 | 199 |
check(mcf.totalCost() == total, "The flow is not optimal " + test_id); |
199 | 200 |
check(checkPotential(gr, lower, upper, cost, mcf.flowMap(), |
200 | 201 |
mcf.potentialMap()), |
201 | 202 |
"Wrong potentials " + test_id); |
202 | 203 |
} |
203 | 204 |
} |
204 | 205 |
|
205 | 206 |
int main() |
206 | 207 |
{ |
207 | 208 |
// Check the interfaces |
208 | 209 |
{ |
209 |
typedef int |
|
210 |
typedef int Flow; |
|
211 |
typedef int Cost; |
|
210 | 212 |
// TODO: This typedef should be enabled if the standard maps are |
211 | 213 |
// reference maps in the graph concepts (See #190). |
212 | 214 |
/**/ |
213 | 215 |
//typedef concepts::Digraph GR; |
214 | 216 |
typedef ListDigraph GR; |
215 | 217 |
/**/ |
216 |
checkConcept< McfClassConcept<GR, Value>, |
|
217 |
NetworkSimplex<GR, Value> >(); |
|
218 |
checkConcept< McfClassConcept<GR, Flow, Cost>, |
|
219 |
NetworkSimplex<GR, Flow, Cost> >(); |
|
218 | 220 |
} |
219 | 221 |
|
220 | 222 |
// Run various MCF tests |
221 | 223 |
typedef ListDigraph Digraph; |
222 | 224 |
DIGRAPH_TYPEDEFS(ListDigraph); |
223 | 225 |
|
224 | 226 |
// Read the test digraph |
225 | 227 |
Digraph gr; |
226 | 228 |
Digraph::ArcMap<int> c(gr), l1(gr), l2(gr), u(gr); |
227 | 229 |
Digraph::NodeMap<int> s1(gr), s2(gr), s3(gr); |
228 | 230 |
ConstMap<Arc, int> cc(1), cu(std::numeric_limits<int>::max()); |
229 | 231 |
Node v, w; |
230 | 232 |
|
231 | 233 |
std::istringstream input(test_lgf); |
232 | 234 |
DigraphReader<Digraph>(gr, input) |
233 | 235 |
.arcMap("cost", c) |
234 | 236 |
.arcMap("cap", u) |
235 | 237 |
.arcMap("low1", l1) |
236 | 238 |
.arcMap("low2", l2) |
237 | 239 |
.nodeMap("sup1", s1) |
238 | 240 |
.nodeMap("sup2", s2) |
239 | 241 |
.nodeMap("sup3", s3) |
240 | 242 |
.node("source", v) |
241 | 243 |
.node("target", w) |
242 | 244 |
.run(); |
243 | 245 |
|
244 | 246 |
// A. Test NetworkSimplex with the default pivot rule |
245 | 247 |
{ |
246 | 248 |
NetworkSimplex<Digraph> mcf(gr); |
247 | 249 |
|
248 | 250 |
mcf.upperMap(u).costMap(c); |
249 | 251 |
checkMcf(mcf, mcf.supplyMap(s1).run(), |
250 | 252 |
gr, l1, u, c, s1, true, 5240, "#A1"); |
251 | 253 |
checkMcf(mcf, mcf.stSupply(v, w, 27).run(), |
252 | 254 |
gr, l1, u, c, s2, true, 7620, "#A2"); |
253 | 255 |
mcf.lowerMap(l2); |
254 | 256 |
checkMcf(mcf, mcf.supplyMap(s1).run(), |
255 | 257 |
gr, l2, u, c, s1, true, 5970, "#A3"); |
256 | 258 |
checkMcf(mcf, mcf.stSupply(v, w, 27).run(), |
257 | 259 |
gr, l2, u, c, s2, true, 8010, "#A4"); |
258 | 260 |
mcf.reset(); |
259 | 261 |
checkMcf(mcf, mcf.supplyMap(s1).run(), |
260 | 262 |
gr, l1, cu, cc, s1, true, 74, "#A5"); |
261 | 263 |
checkMcf(mcf, mcf.lowerMap(l2).stSupply(v, w, 27).run(), |
262 | 264 |
gr, l2, cu, cc, s2, true, 94, "#A6"); |
263 | 265 |
mcf.reset(); |
264 | 266 |
checkMcf(mcf, mcf.run(), |
265 | 267 |
gr, l1, cu, cc, s3, true, 0, "#A7"); |
266 | 268 |
checkMcf(mcf, mcf.boundMaps(l2, u).run(), |
267 | 269 |
gr, l2, u, cc, s3, false, 0, "#A8"); |
268 | 270 |
} |
269 | 271 |
|
270 | 272 |
// B. Test NetworkSimplex with each pivot rule |
271 | 273 |
{ |
272 | 274 |
NetworkSimplex<Digraph> mcf(gr); |
273 | 275 |
mcf.supplyMap(s1).costMap(c).capacityMap(u).lowerMap(l2); |
274 | 276 |
|
275 | 277 |
checkMcf(mcf, mcf.run(NetworkSimplex<Digraph>::FIRST_ELIGIBLE), |
276 | 278 |
gr, l2, u, c, s1, true, 5970, "#B1"); |
277 | 279 |
checkMcf(mcf, mcf.run(NetworkSimplex<Digraph>::BEST_ELIGIBLE), |
278 | 280 |
gr, l2, u, c, s1, true, 5970, "#B2"); |
279 | 281 |
checkMcf(mcf, mcf.run(NetworkSimplex<Digraph>::BLOCK_SEARCH), |
280 | 282 |
gr, l2, u, c, s1, true, 5970, "#B3"); |
281 | 283 |
checkMcf(mcf, mcf.run(NetworkSimplex<Digraph>::CANDIDATE_LIST), |
282 | 284 |
gr, l2, u, c, s1, true, 5970, "#B4"); |
283 | 285 |
checkMcf(mcf, mcf.run(NetworkSimplex<Digraph>::ALTERING_LIST), |
284 | 286 |
gr, l2, u, c, s1, true, 5970, "#B5"); |
285 | 287 |
} |
286 | 288 |
|
287 | 289 |
return 0; |
288 | 290 |
} |
0 comments (0 inline)