1 /* -*- mode: C++; indent-tabs-mode: nil; -*-
3 * This file is a part of LEMON, a generic C++ optimization library.
5 * Copyright (C) 2003-2010
6 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 * (Egervary Research Group on Combinatorial Optimization, EGRES).
9 * Permission to use, modify and distribute this software is granted
10 * provided that this copyright notice appears in all copies. For
11 * precise terms see the accompanying LICENSE file.
13 * This software is provided "AS IS" with no warranty of any kind,
14 * express or implied, and with no claim as to its suitability for any
19 #ifndef LEMON_HARTMANN_ORLIN_MMC_H
20 #define LEMON_HARTMANN_ORLIN_MMC_H
22 /// \ingroup min_mean_cycle
25 /// \brief Hartmann-Orlin's algorithm for finding a minimum mean cycle.
29 #include <lemon/core.h>
30 #include <lemon/path.h>
31 #include <lemon/tolerance.h>
32 #include <lemon/connectivity.h>
36 /// \brief Default traits class of HartmannOrlinMmc class.
38 /// Default traits class of HartmannOrlinMmc class.
39 /// \tparam GR The type of the digraph.
40 /// \tparam CM The type of the cost map.
41 /// It must conform to the \ref concepts::ReadMap "ReadMap" concept.
43 template <typename GR, typename CM>
45 template <typename GR, typename CM,
46 bool integer = std::numeric_limits<typename CM::Value>::is_integer>
48 struct HartmannOrlinMmcDefaultTraits
50 /// The type of the digraph
52 /// The type of the cost map
54 /// The type of the arc costs
55 typedef typename CostMap::Value Cost;
57 /// \brief The large cost type used for internal computations
59 /// The large cost type used for internal computations.
60 /// It is \c long \c long if the \c Cost type is integer,
61 /// otherwise it is \c double.
62 /// \c Cost must be convertible to \c LargeCost.
63 typedef double LargeCost;
65 /// The tolerance type used for internal computations
66 typedef lemon::Tolerance<LargeCost> Tolerance;
68 /// \brief The path type of the found cycles
70 /// The path type of the found cycles.
71 /// It must conform to the \ref lemon::concepts::Path "Path" concept
72 /// and it must have an \c addFront() function.
73 typedef lemon::Path<Digraph> Path;
76 // Default traits class for integer cost types
77 template <typename GR, typename CM>
78 struct HartmannOrlinMmcDefaultTraits<GR, CM, true>
82 typedef typename CostMap::Value Cost;
83 #ifdef LEMON_HAVE_LONG_LONG
84 typedef long long LargeCost;
86 typedef long LargeCost;
88 typedef lemon::Tolerance<LargeCost> Tolerance;
89 typedef lemon::Path<Digraph> Path;
93 /// \addtogroup min_mean_cycle
96 /// \brief Implementation of the Hartmann-Orlin algorithm for finding
97 /// a minimum mean cycle.
99 /// This class implements the Hartmann-Orlin algorithm for finding
100 /// a directed cycle of minimum mean cost in a digraph
101 /// \ref hartmann93finding, \ref dasdan98minmeancycle.
102 /// This method is based on \ref KarpMmc "Karp"'s original algorithm, but
103 /// applies an early termination scheme. It makes the algorithm
104 /// significantly faster for some problem instances, but slower for others.
105 /// The algorithm runs in time O(ne) and uses space O(n<sup>2</sup>+e).
107 /// \tparam GR The type of the digraph the algorithm runs on.
108 /// \tparam CM The type of the cost map. The default
109 /// map type is \ref concepts::Digraph::ArcMap "GR::ArcMap<int>".
110 /// \tparam TR The traits class that defines various types used by the
111 /// algorithm. By default, it is \ref HartmannOrlinMmcDefaultTraits
112 /// "HartmannOrlinMmcDefaultTraits<GR, CM>".
113 /// In most cases, this parameter should not be set directly,
114 /// consider to use the named template parameters instead.
116 template <typename GR, typename CM, typename TR>
118 template < typename GR,
119 typename CM = typename GR::template ArcMap<int>,
120 typename TR = HartmannOrlinMmcDefaultTraits<GR, CM> >
122 class HartmannOrlinMmc
126 /// The type of the digraph
127 typedef typename TR::Digraph Digraph;
128 /// The type of the cost map
129 typedef typename TR::CostMap CostMap;
130 /// The type of the arc costs
131 typedef typename TR::Cost Cost;
133 /// \brief The large cost type
135 /// The large cost type used for internal computations.
136 /// By default, it is \c long \c long if the \c Cost type is integer,
137 /// otherwise it is \c double.
138 typedef typename TR::LargeCost LargeCost;
140 /// The tolerance type
141 typedef typename TR::Tolerance Tolerance;
143 /// \brief The path type of the found cycles
145 /// The path type of the found cycles.
146 /// Using the \ref HartmannOrlinMmcDefaultTraits "default traits class",
147 /// it is \ref lemon::Path "Path<Digraph>".
148 typedef typename TR::Path Path;
150 /// The \ref HartmannOrlinMmcDefaultTraits "traits class" of the algorithm
155 TEMPLATE_DIGRAPH_TYPEDEFS(Digraph);
157 // Data sturcture for path data
162 PathData(LargeCost d, Arc p = INVALID) :
166 typedef typename Digraph::template NodeMap<std::vector<PathData> >
171 // The digraph the algorithm runs on
173 // The cost of the arcs
174 const CostMap &_cost;
176 // Data for storing the strongly connected components
178 typename Digraph::template NodeMap<int> _comp;
179 std::vector<std::vector<Node> > _comp_nodes;
180 std::vector<Node>* _nodes;
181 typename Digraph::template NodeMap<std::vector<Arc> > _out_arcs;
183 // Data for the found cycles
184 bool _curr_found, _best_found;
185 LargeCost _curr_cost, _best_cost;
186 int _curr_size, _best_size;
187 Node _curr_node, _best_node;
188 int _curr_level, _best_level;
193 // Node map for storing path data
194 PathDataNodeMap _data;
195 // The processed nodes in the last round
196 std::vector<Node> _process;
198 Tolerance _tolerance;
205 /// \name Named Template Parameters
208 template <typename T>
209 struct SetLargeCostTraits : public Traits {
211 typedef lemon::Tolerance<T> Tolerance;
214 /// \brief \ref named-templ-param "Named parameter" for setting
215 /// \c LargeCost type.
217 /// \ref named-templ-param "Named parameter" for setting \c LargeCost
218 /// type. It is used for internal computations in the algorithm.
219 template <typename T>
221 : public HartmannOrlinMmc<GR, CM, SetLargeCostTraits<T> > {
222 typedef HartmannOrlinMmc<GR, CM, SetLargeCostTraits<T> > Create;
225 template <typename T>
226 struct SetPathTraits : public Traits {
230 /// \brief \ref named-templ-param "Named parameter" for setting
233 /// \ref named-templ-param "Named parameter" for setting the \c %Path
234 /// type of the found cycles.
235 /// It must conform to the \ref lemon::concepts::Path "Path" concept
236 /// and it must have an \c addFront() function.
237 template <typename T>
239 : public HartmannOrlinMmc<GR, CM, SetPathTraits<T> > {
240 typedef HartmannOrlinMmc<GR, CM, SetPathTraits<T> > Create;
247 HartmannOrlinMmc() {}
251 /// \brief Constructor.
253 /// The constructor of the class.
255 /// \param digraph The digraph the algorithm runs on.
256 /// \param cost The costs of the arcs.
257 HartmannOrlinMmc( const Digraph &digraph,
258 const CostMap &cost ) :
259 _gr(digraph), _cost(cost), _comp(digraph), _out_arcs(digraph),
260 _best_found(false), _best_cost(0), _best_size(1),
261 _cycle_path(NULL), _local_path(false), _data(digraph),
262 INF(std::numeric_limits<LargeCost>::has_infinity ?
263 std::numeric_limits<LargeCost>::infinity() :
264 std::numeric_limits<LargeCost>::max())
268 ~HartmannOrlinMmc() {
269 if (_local_path) delete _cycle_path;
272 /// \brief Set the path structure for storing the found cycle.
274 /// This function sets an external path structure for storing the
277 /// If you don't call this function before calling \ref run() or
278 /// \ref findCycleMean(), a local \ref Path "path" structure
279 /// will be allocated. The destuctor deallocates this automatically
280 /// allocated object, of course.
282 /// \note The algorithm calls only the \ref lemon::Path::addFront()
283 /// "addFront()" function of the given path structure.
285 /// \return <tt>(*this)</tt>
286 HartmannOrlinMmc& cycle(Path &path) {
295 /// \brief Set the tolerance used by the algorithm.
297 /// This function sets the tolerance object used by the algorithm.
299 /// \return <tt>(*this)</tt>
300 HartmannOrlinMmc& tolerance(const Tolerance& tolerance) {
301 _tolerance = tolerance;
305 /// \brief Return a const reference to the tolerance.
307 /// This function returns a const reference to the tolerance object
308 /// used by the algorithm.
309 const Tolerance& tolerance() const {
313 /// \name Execution control
314 /// The simplest way to execute the algorithm is to call the \ref run()
316 /// If you only need the minimum mean cost, you may call
317 /// \ref findCycleMean().
321 /// \brief Run the algorithm.
323 /// This function runs the algorithm.
324 /// It can be called more than once (e.g. if the underlying digraph
325 /// and/or the arc costs have been modified).
327 /// \return \c true if a directed cycle exists in the digraph.
329 /// \note <tt>mmc.run()</tt> is just a shortcut of the following code.
331 /// return mmc.findCycleMean() && mmc.findCycle();
334 return findCycleMean() && findCycle();
337 /// \brief Find the minimum cycle mean.
339 /// This function finds the minimum mean cost of the directed
340 /// cycles in the digraph.
342 /// \return \c true if a directed cycle exists in the digraph.
343 bool findCycleMean() {
344 // Initialization and find strongly connected components
348 // Find the minimum cycle mean in the components
349 for (int comp = 0; comp < _comp_num; ++comp) {
350 if (!initComponent(comp)) continue;
353 // Update the best cycle (global minimum mean cycle)
354 if ( _curr_found && (!_best_found ||
355 _curr_cost * _best_size < _best_cost * _curr_size) ) {
357 _best_cost = _curr_cost;
358 _best_size = _curr_size;
359 _best_node = _curr_node;
360 _best_level = _curr_level;
366 /// \brief Find a minimum mean directed cycle.
368 /// This function finds a directed cycle of minimum mean cost
369 /// in the digraph using the data computed by findCycleMean().
371 /// \return \c true if a directed cycle exists in the digraph.
373 /// \pre \ref findCycleMean() must be called before using this function.
375 if (!_best_found) return false;
376 IntNodeMap reached(_gr, -1);
377 int r = _best_level + 1;
379 while (reached[u] < 0) {
381 u = _gr.source(_data[u][r].pred);
384 Arc e = _data[u][r].pred;
385 _cycle_path->addFront(e);
386 _best_cost = _cost[e];
389 while ((v = _gr.source(e)) != u) {
390 e = _data[v][--r].pred;
391 _cycle_path->addFront(e);
392 _best_cost += _cost[e];
400 /// \name Query Functions
401 /// The results of the algorithm can be obtained using these
403 /// The algorithm should be executed before using them.
407 /// \brief Return the total cost of the found cycle.
409 /// This function returns the total cost of the found cycle.
411 /// \pre \ref run() or \ref findCycleMean() must be called before
412 /// using this function.
413 Cost cycleCost() const {
414 return static_cast<Cost>(_best_cost);
417 /// \brief Return the number of arcs on the found cycle.
419 /// This function returns the number of arcs on the found cycle.
421 /// \pre \ref run() or \ref findCycleMean() must be called before
422 /// using this function.
423 int cycleSize() const {
427 /// \brief Return the mean cost of the found cycle.
429 /// This function returns the mean cost of the found cycle.
431 /// \note <tt>alg.cycleMean()</tt> is just a shortcut of the
434 /// return static_cast<double>(alg.cycleCost()) / alg.cycleSize();
437 /// \pre \ref run() or \ref findCycleMean() must be called before
438 /// using this function.
439 double cycleMean() const {
440 return static_cast<double>(_best_cost) / _best_size;
443 /// \brief Return the found cycle.
445 /// This function returns a const reference to the path structure
446 /// storing the found cycle.
448 /// \pre \ref run() or \ref findCycle() must be called before using
450 const Path& cycle() const {
462 _cycle_path = new Path;
464 _cycle_path->clear();
468 _cycle_path->clear();
469 for (NodeIt u(_gr); u != INVALID; ++u)
473 // Find strongly connected components and initialize _comp_nodes
475 void findComponents() {
476 _comp_num = stronglyConnectedComponents(_gr, _comp);
477 _comp_nodes.resize(_comp_num);
478 if (_comp_num == 1) {
479 _comp_nodes[0].clear();
480 for (NodeIt n(_gr); n != INVALID; ++n) {
481 _comp_nodes[0].push_back(n);
482 _out_arcs[n].clear();
483 for (OutArcIt a(_gr, n); a != INVALID; ++a) {
484 _out_arcs[n].push_back(a);
488 for (int i = 0; i < _comp_num; ++i)
489 _comp_nodes[i].clear();
490 for (NodeIt n(_gr); n != INVALID; ++n) {
492 _comp_nodes[k].push_back(n);
493 _out_arcs[n].clear();
494 for (OutArcIt a(_gr, n); a != INVALID; ++a) {
495 if (_comp[_gr.target(a)] == k) _out_arcs[n].push_back(a);
501 // Initialize path data for the current component
502 bool initComponent(int comp) {
503 _nodes = &(_comp_nodes[comp]);
504 int n = _nodes->size();
505 if (n < 1 || (n == 1 && _out_arcs[(*_nodes)[0]].size() == 0)) {
508 for (int i = 0; i < n; ++i) {
509 _data[(*_nodes)[i]].resize(n + 1, PathData(INF));
514 // Process all rounds of computing path data for the current component.
515 // _data[v][k] is the cost of a shortest directed walk from the root
516 // node to node v containing exactly k arcs.
517 void processRounds() {
518 Node start = (*_nodes)[0];
519 _data[start][0] = PathData(0);
521 _process.push_back(start);
523 int k, n = _nodes->size();
525 bool terminate = false;
526 for (k = 1; k <= n && int(_process.size()) < n && !terminate; ++k) {
527 processNextBuildRound(k);
528 if (k == next_check || k == n) {
529 terminate = checkTermination(k);
530 next_check = next_check * 3 / 2;
533 for ( ; k <= n && !terminate; ++k) {
534 processNextFullRound(k);
535 if (k == next_check || k == n) {
536 terminate = checkTermination(k);
537 next_check = next_check * 3 / 2;
542 // Process one round and rebuild _process
543 void processNextBuildRound(int k) {
544 std::vector<Node> next;
548 for (int i = 0; i < int(_process.size()); ++i) {
550 for (int j = 0; j < int(_out_arcs[u].size()); ++j) {
553 d = _data[u][k-1].dist + _cost[e];
554 if (_tolerance.less(d, _data[v][k].dist)) {
555 if (_data[v][k].dist == INF) next.push_back(v);
556 _data[v][k] = PathData(d, e);
563 // Process one round using _nodes instead of _process
564 void processNextFullRound(int k) {
568 for (int i = 0; i < int(_nodes->size()); ++i) {
570 for (int j = 0; j < int(_out_arcs[u].size()); ++j) {
573 d = _data[u][k-1].dist + _cost[e];
574 if (_tolerance.less(d, _data[v][k].dist)) {
575 _data[v][k] = PathData(d, e);
581 // Check early termination
582 bool checkTermination(int k) {
583 typedef std::pair<int, int> Pair;
584 typename GR::template NodeMap<Pair> level(_gr, Pair(-1, 0));
585 typename GR::template NodeMap<LargeCost> pi(_gr);
586 int n = _nodes->size();
591 // Search for cycles that are already found
593 for (int i = 0; i < n; ++i) {
595 if (_data[u][k].dist == INF) continue;
596 for (int j = k; j >= 0; --j) {
597 if (level[u].first == i && level[u].second > 0) {
599 cost = _data[u][level[u].second].dist - _data[u][j].dist;
600 size = level[u].second - j;
601 if (!_curr_found || cost * _curr_size < _curr_cost * size) {
605 _curr_level = level[u].second;
609 level[u] = Pair(i, j);
611 u = _gr.source(_data[u][j].pred);
616 // If at least one cycle is found, check the optimality condition
618 if (_curr_found && k < n) {
619 // Find node potentials
620 for (int i = 0; i < n; ++i) {
623 for (int j = 0; j <= k; ++j) {
624 if (_data[u][j].dist < INF) {
625 d = _data[u][j].dist * _curr_size - j * _curr_cost;
626 if (_tolerance.less(d, pi[u])) pi[u] = d;
631 // Check the optimality condition for all arcs
633 for (ArcIt a(_gr); a != INVALID; ++a) {
634 if (_tolerance.less(_cost[a] * _curr_size - _curr_cost,
635 pi[_gr.target(a)] - pi[_gr.source(a)]) ) {
645 }; //class HartmannOrlinMmc
651 #endif //LEMON_HARTMANN_ORLIN_MMC_H