lemon/capacity_scaling.h
author kpeter
Mon, 18 Feb 2008 03:30:53 +0000
changeset 2574 7058c9690e7d
parent 2556 74c2c81055e1
child 2581 054566ac0934
permissions -rw-r--r--
Improvements in CapacityScaling.

Main changes:
- Use -potenital[] instead of potential[] to conform to the usual
terminology.
- Change the name of private members to start with "_".
- Change the name of function parameters not to start with "_".
- Remove unnecessary documentation for private members.
- Doc improvements.
     1 /* -*- C++ -*-
     2  *
     3  * This file is a part of LEMON, a generic C++ optimization library
     4  *
     5  * Copyright (C) 2003-2008
     6  * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
     7  * (Egervary Research Group on Combinatorial Optimization, EGRES).
     8  *
     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.
    12  *
    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
    15  * purpose.
    16  *
    17  */
    18 
    19 #ifndef LEMON_CAPACITY_SCALING_H
    20 #define LEMON_CAPACITY_SCALING_H
    21 
    22 /// \ingroup min_cost_flow
    23 ///
    24 /// \file
    25 /// \brief Capacity scaling algorithm for finding a minimum cost flow.
    26 
    27 #include <vector>
    28 
    29 #include <lemon/graph_adaptor.h>
    30 #include <lemon/bin_heap.h>
    31 
    32 namespace lemon {
    33 
    34   /// \addtogroup min_cost_flow
    35   /// @{
    36 
    37   /// \brief Implementation of the capacity scaling algorithm for
    38   /// finding a minimum cost flow.
    39   ///
    40   /// \ref CapacityScaling implements the capacity scaling version
    41   /// of the successive shortest path algorithm for finding a minimum
    42   /// cost flow.
    43   ///
    44   /// \tparam Graph The directed graph type the algorithm runs on.
    45   /// \tparam LowerMap The type of the lower bound map.
    46   /// \tparam CapacityMap The type of the capacity (upper bound) map.
    47   /// \tparam CostMap The type of the cost (length) map.
    48   /// \tparam SupplyMap The type of the supply map.
    49   ///
    50   /// \warning
    51   /// - Edge capacities and costs should be \e non-negative \e integers.
    52   /// - Supply values should be \e signed \e integers.
    53   /// - \c LowerMap::Value must be convertible to \c CapacityMap::Value.
    54   /// - \c CapacityMap::Value and \c SupplyMap::Value must be
    55   ///   convertible to each other.
    56   /// - All value types must be convertible to \c CostMap::Value, which
    57   ///   must be signed type.
    58   ///
    59   /// \author Peter Kovacs
    60 
    61   template < typename Graph,
    62              typename LowerMap = typename Graph::template EdgeMap<int>,
    63              typename CapacityMap = typename Graph::template EdgeMap<int>,
    64              typename CostMap = typename Graph::template EdgeMap<int>,
    65              typename SupplyMap = typename Graph::template NodeMap<int> >
    66   class CapacityScaling
    67   {
    68     GRAPH_TYPEDEFS(typename Graph);
    69 
    70     typedef typename CapacityMap::Value Capacity;
    71     typedef typename CostMap::Value Cost;
    72     typedef typename SupplyMap::Value Supply;
    73     typedef typename Graph::template EdgeMap<Capacity> CapacityEdgeMap;
    74     typedef typename Graph::template NodeMap<Supply> SupplyNodeMap;
    75     typedef typename Graph::template NodeMap<Edge> PredMap;
    76 
    77   public:
    78 
    79     /// The type of the flow map.
    80     typedef typename Graph::template EdgeMap<Capacity> FlowMap;
    81     /// The type of the potential map.
    82     typedef typename Graph::template NodeMap<Cost> PotentialMap;
    83 
    84   private:
    85 
    86     /// \brief Special implementation of the \ref Dijkstra algorithm
    87     /// for finding shortest paths in the residual network.
    88     ///
    89     /// \ref ResidualDijkstra is a special implementation of the
    90     /// \ref Dijkstra algorithm for finding shortest paths in the
    91     /// residual network of the graph with respect to the reduced edge
    92     /// costs and modifying the node potentials according to the
    93     /// distance of the nodes.
    94     class ResidualDijkstra
    95     {
    96       typedef typename Graph::template NodeMap<Cost> CostNodeMap;
    97       typedef typename Graph::template NodeMap<Edge> PredMap;
    98 
    99       typedef typename Graph::template NodeMap<int> HeapCrossRef;
   100       typedef BinHeap<Cost, HeapCrossRef> Heap;
   101 
   102     private:
   103 
   104       // The directed graph the algorithm runs on
   105       const Graph &_graph;
   106 
   107       // The main maps
   108       const FlowMap &_flow;
   109       const CapacityEdgeMap &_res_cap;
   110       const CostMap &_cost;
   111       const SupplyNodeMap &_excess;
   112       PotentialMap &_potential;
   113 
   114       // The distance map
   115       CostNodeMap _dist;
   116       // The pred edge map
   117       PredMap &_pred;
   118       // The processed (i.e. permanently labeled) nodes
   119       std::vector<Node> _proc_nodes;
   120 
   121     public:
   122 
   123       /// The constructor of the class.
   124       ResidualDijkstra( const Graph &graph,
   125                         const FlowMap &flow,
   126                         const CapacityEdgeMap &res_cap,
   127                         const CostMap &cost,
   128                         const SupplyMap &excess,
   129                         PotentialMap &potential,
   130                         PredMap &pred ) :
   131         _graph(graph), _flow(flow), _res_cap(res_cap), _cost(cost),
   132         _excess(excess), _potential(potential), _dist(graph),
   133         _pred(pred)
   134       {}
   135 
   136       /// Runs the algorithm from the given source node.
   137       Node run(Node s, Capacity delta) {
   138         HeapCrossRef heap_cross_ref(_graph, Heap::PRE_HEAP);
   139         Heap heap(heap_cross_ref);
   140         heap.push(s, 0);
   141         _pred[s] = INVALID;
   142         _proc_nodes.clear();
   143 
   144         // Processing nodes
   145         while (!heap.empty() && _excess[heap.top()] > -delta) {
   146           Node u = heap.top(), v;
   147           Cost d = heap.prio() + _potential[u], nd;
   148           _dist[u] = heap.prio();
   149           heap.pop();
   150           _proc_nodes.push_back(u);
   151 
   152           // Traversing outgoing edges
   153           for (OutEdgeIt e(_graph, u); e != INVALID; ++e) {
   154             if (_res_cap[e] >= delta) {
   155               v = _graph.target(e);
   156               switch(heap.state(v)) {
   157               case Heap::PRE_HEAP:
   158                 heap.push(v, d + _cost[e] - _potential[v]);
   159                 _pred[v] = e;
   160                 break;
   161               case Heap::IN_HEAP:
   162                 nd = d + _cost[e] - _potential[v];
   163                 if (nd < heap[v]) {
   164                   heap.decrease(v, nd);
   165                   _pred[v] = e;
   166                 }
   167                 break;
   168               case Heap::POST_HEAP:
   169                 break;
   170               }
   171             }
   172           }
   173 
   174           // Traversing incoming edges
   175           for (InEdgeIt e(_graph, u); e != INVALID; ++e) {
   176             if (_flow[e] >= delta) {
   177               v = _graph.source(e);
   178               switch(heap.state(v)) {
   179               case Heap::PRE_HEAP:
   180                 heap.push(v, d - _cost[e] - _potential[v]);
   181                 _pred[v] = e;
   182                 break;
   183               case Heap::IN_HEAP:
   184                 nd = d - _cost[e] - _potential[v];
   185                 if (nd < heap[v]) {
   186                   heap.decrease(v, nd);
   187                   _pred[v] = e;
   188                 }
   189                 break;
   190               case Heap::POST_HEAP:
   191                 break;
   192               }
   193             }
   194           }
   195         }
   196         if (heap.empty()) return INVALID;
   197 
   198         // Updating potentials of processed nodes
   199         Node t = heap.top();
   200         Cost t_dist = heap.prio();
   201         for (int i = 0; i < int(_proc_nodes.size()); ++i)
   202           _potential[_proc_nodes[i]] += _dist[_proc_nodes[i]] - t_dist;
   203 
   204         return t;
   205       }
   206 
   207     }; //class ResidualDijkstra
   208 
   209   private:
   210 
   211     // The directed graph the algorithm runs on
   212     const Graph &_graph;
   213     // The original lower bound map
   214     const LowerMap *_lower;
   215     // The modified capacity map
   216     CapacityEdgeMap _capacity;
   217     // The original cost map
   218     const CostMap &_cost;
   219     // The modified supply map
   220     SupplyNodeMap _supply;
   221     bool _valid_supply;
   222 
   223     // Edge map of the current flow
   224     FlowMap _flow;
   225     // Node map of the current potentials
   226     PotentialMap _potential;
   227 
   228     // The residual capacity map
   229     CapacityEdgeMap _res_cap;
   230     // The excess map
   231     SupplyNodeMap _excess;
   232     // The excess nodes (i.e. nodes with positive excess)
   233     std::vector<Node> _excess_nodes;
   234     // The deficit nodes (i.e. nodes with negative excess)
   235     std::vector<Node> _deficit_nodes;
   236 
   237     // The delta parameter used for capacity scaling
   238     Capacity _delta;
   239     // The maximum number of phases
   240     int _phase_num;
   241 
   242     // The pred edge map
   243     PredMap _pred;
   244     // Implementation of the Dijkstra algorithm for finding augmenting
   245     // shortest paths in the residual network
   246     ResidualDijkstra _dijkstra;
   247 
   248   public :
   249 
   250     /// \brief General constructor of the class (with lower bounds).
   251     ///
   252     /// General constructor of the class (with lower bounds).
   253     ///
   254     /// \param graph The directed graph the algorithm runs on.
   255     /// \param lower The lower bounds of the edges.
   256     /// \param capacity The capacities (upper bounds) of the edges.
   257     /// \param cost The cost (length) values of the edges.
   258     /// \param supply The supply values of the nodes (signed).
   259     CapacityScaling( const Graph &graph,
   260                      const LowerMap &lower,
   261                      const CapacityMap &capacity,
   262                      const CostMap &cost,
   263                      const SupplyMap &supply ) :
   264       _graph(graph), _lower(&lower), _capacity(graph), _cost(cost),
   265       _supply(graph), _flow(graph, 0), _potential(graph, 0),
   266       _res_cap(graph), _excess(graph), _pred(graph),
   267       _dijkstra(_graph, _flow, _res_cap, _cost, _excess, _potential, _pred)
   268     {
   269       // Removing non-zero lower bounds
   270       _capacity = subMap(capacity, lower);
   271       _res_cap = _capacity;
   272       Supply sum = 0;
   273       for (NodeIt n(_graph); n != INVALID; ++n) {
   274         Supply s = supply[n];
   275         for (InEdgeIt e(_graph, n); e != INVALID; ++e)
   276           s += lower[e];
   277         for (OutEdgeIt e(_graph, n); e != INVALID; ++e)
   278           s -= lower[e];
   279         _supply[n] = s;
   280         sum += s;
   281       }
   282       _valid_supply = sum == 0;
   283     }
   284 
   285     /// \brief General constructor of the class (without lower bounds).
   286     ///
   287     /// General constructor of the class (without lower bounds).
   288     ///
   289     /// \param graph The directed graph the algorithm runs on.
   290     /// \param capacity The capacities (upper bounds) of the edges.
   291     /// \param cost The cost (length) values of the edges.
   292     /// \param supply The supply values of the nodes (signed).
   293     CapacityScaling( const Graph &graph,
   294                      const CapacityMap &capacity,
   295                      const CostMap &cost,
   296                      const SupplyMap &supply ) :
   297       _graph(graph), _lower(NULL), _capacity(capacity), _cost(cost),
   298       _supply(supply), _flow(graph, 0), _potential(graph, 0),
   299       _res_cap(capacity), _excess(graph), _pred(graph),
   300       _dijkstra(_graph, _flow, _res_cap, _cost, _excess, _potential, _pred)
   301     {
   302       // Checking the sum of supply values
   303       Supply sum = 0;
   304       for (NodeIt n(_graph); n != INVALID; ++n) sum += _supply[n];
   305       _valid_supply = sum == 0;
   306     }
   307 
   308     /// \brief Simple constructor of the class (with lower bounds).
   309     ///
   310     /// Simple constructor of the class (with lower bounds).
   311     ///
   312     /// \param graph The directed graph the algorithm runs on.
   313     /// \param lower The lower bounds of the edges.
   314     /// \param capacity The capacities (upper bounds) of the edges.
   315     /// \param cost The cost (length) values of the edges.
   316     /// \param s The source node.
   317     /// \param t The target node.
   318     /// \param flow_value The required amount of flow from node \c s
   319     /// to node \c t (i.e. the supply of \c s and the demand of \c t).
   320     CapacityScaling( const Graph &graph,
   321                      const LowerMap &lower,
   322                      const CapacityMap &capacity,
   323                      const CostMap &cost,
   324                      Node s, Node t,
   325                      Supply flow_value ) :
   326       _graph(graph), _lower(&lower), _capacity(graph), _cost(cost),
   327       _supply(graph), _flow(graph, 0), _potential(graph, 0),
   328       _res_cap(graph), _excess(graph), _pred(graph),
   329       _dijkstra(_graph, _flow, _res_cap, _cost, _excess, _potential, _pred)
   330     {
   331       // Removing non-zero lower bounds
   332       _capacity = subMap(capacity, lower);
   333       _res_cap = _capacity;
   334       for (NodeIt n(_graph); n != INVALID; ++n) {
   335         Supply sum = 0;
   336         if (n == s) sum =  flow_value;
   337         if (n == t) sum = -flow_value;
   338         for (InEdgeIt e(_graph, n); e != INVALID; ++e)
   339           sum += lower[e];
   340         for (OutEdgeIt e(_graph, n); e != INVALID; ++e)
   341           sum -= lower[e];
   342         _supply[n] = sum;
   343       }
   344       _valid_supply = true;
   345     }
   346 
   347     /// \brief Simple constructor of the class (without lower bounds).
   348     ///
   349     /// Simple constructor of the class (without lower bounds).
   350     ///
   351     /// \param graph The directed graph the algorithm runs on.
   352     /// \param capacity The capacities (upper bounds) of the edges.
   353     /// \param cost The cost (length) values of the edges.
   354     /// \param s The source node.
   355     /// \param t The target node.
   356     /// \param flow_value The required amount of flow from node \c s
   357     /// to node \c t (i.e. the supply of \c s and the demand of \c t).
   358     CapacityScaling( const Graph &graph,
   359                      const CapacityMap &capacity,
   360                      const CostMap &cost,
   361                      Node s, Node t,
   362                      Supply flow_value ) :
   363       _graph(graph), _lower(NULL), _capacity(capacity), _cost(cost),
   364       _supply(graph, 0), _flow(graph, 0), _potential(graph, 0),
   365       _res_cap(capacity), _excess(graph), _pred(graph),
   366       _dijkstra(_graph, _flow, _res_cap, _cost, _excess, _potential, _pred)
   367     {
   368       _supply[s] =  flow_value;
   369       _supply[t] = -flow_value;
   370       _valid_supply = true;
   371     }
   372 
   373     /// \brief Runs the algorithm.
   374     ///
   375     /// Runs the algorithm.
   376     ///
   377     /// \param scaling Enable or disable capacity scaling.
   378     /// If the maximum edge capacity and/or the amount of total supply
   379     /// is rather small, the algorithm could be slightly faster without
   380     /// scaling.
   381     ///
   382     /// \return \c true if a feasible flow can be found.
   383     bool run(bool scaling = true) {
   384       return init(scaling) && start();
   385     }
   386 
   387     /// \brief Returns a const reference to the edge map storing the
   388     /// found flow.
   389     ///
   390     /// Returns a const reference to the edge map storing the found flow.
   391     ///
   392     /// \pre \ref run() must be called before using this function.
   393     const FlowMap& flowMap() const {
   394       return _flow;
   395     }
   396 
   397     /// \brief Returns a const reference to the node map storing the
   398     /// found potentials (the dual solution).
   399     ///
   400     /// Returns a const reference to the node map storing the found
   401     /// potentials (the dual solution).
   402     ///
   403     /// \pre \ref run() must be called before using this function.
   404     const PotentialMap& potentialMap() const {
   405       return _potential;
   406     }
   407 
   408     /// \brief Returns the total cost of the found flow.
   409     ///
   410     /// Returns the total cost of the found flow. The complexity of the
   411     /// function is \f$ O(e) \f$.
   412     ///
   413     /// \pre \ref run() must be called before using this function.
   414     Cost totalCost() const {
   415       Cost c = 0;
   416       for (EdgeIt e(_graph); e != INVALID; ++e)
   417         c += _flow[e] * _cost[e];
   418       return c;
   419     }
   420 
   421   private:
   422 
   423     /// Initializes the algorithm.
   424     bool init(bool scaling) {
   425       if (!_valid_supply) return false;
   426       _excess = _supply;
   427 
   428       // Initilaizing delta value
   429       if (scaling) {
   430         // With scaling
   431         Supply max_sup = 0, max_dem = 0;
   432         for (NodeIt n(_graph); n != INVALID; ++n) {
   433           if ( _supply[n] > max_sup) max_sup =  _supply[n];
   434           if (-_supply[n] > max_dem) max_dem = -_supply[n];
   435         }
   436         if (max_dem < max_sup) max_sup = max_dem;
   437         _phase_num = 0;
   438         for (_delta = 1; 2 * _delta <= max_sup; _delta *= 2)
   439           ++_phase_num;
   440       } else {
   441         // Without scaling
   442         _delta = 1;
   443       }
   444       return true;
   445     }
   446 
   447     bool start() {
   448       if (_delta > 1)
   449         return startWithScaling();
   450       else
   451         return startWithoutScaling();
   452     }
   453 
   454     /// Executes the capacity scaling algorithm.
   455     bool startWithScaling() {
   456       // Processing capacity scaling phases
   457       Node s, t;
   458       int phase_cnt = 0;
   459       int factor = 4;
   460       while (true) {
   461         // Saturating all edges not satisfying the optimality condition
   462         for (EdgeIt e(_graph); e != INVALID; ++e) {
   463           Node u = _graph.source(e), v = _graph.target(e);
   464           Cost c = _cost[e] + _potential[u] - _potential[v];
   465           if (c < 0 && _res_cap[e] >= _delta) {
   466             _excess[u] -= _res_cap[e];
   467             _excess[v] += _res_cap[e];
   468             _flow[e] = _capacity[e];
   469             _res_cap[e] = 0;
   470           }
   471           else if (c > 0 && _flow[e] >= _delta) {
   472             _excess[u] += _flow[e];
   473             _excess[v] -= _flow[e];
   474             _flow[e] = 0;
   475             _res_cap[e] = _capacity[e];
   476           }
   477         }
   478 
   479         // Finding excess nodes and deficit nodes
   480         _excess_nodes.clear();
   481         _deficit_nodes.clear();
   482         for (NodeIt n(_graph); n != INVALID; ++n) {
   483           if (_excess[n] >=  _delta) _excess_nodes.push_back(n);
   484           if (_excess[n] <= -_delta) _deficit_nodes.push_back(n);
   485         }
   486         int next_node = 0;
   487 
   488         // Finding augmenting shortest paths
   489         while (next_node < int(_excess_nodes.size())) {
   490           // Checking deficit nodes
   491           if (_delta > 1) {
   492             bool delta_deficit = false;
   493             for (int i = 0; i < int(_deficit_nodes.size()); ++i) {
   494               if (_excess[_deficit_nodes[i]] <= -_delta) {
   495                 delta_deficit = true;
   496                 break;
   497               }
   498             }
   499             if (!delta_deficit) break;
   500           }
   501 
   502           // Running Dijkstra
   503           s = _excess_nodes[next_node];
   504           if ((t = _dijkstra.run(s, _delta)) == INVALID) {
   505             if (_delta > 1) {
   506               ++next_node;
   507               continue;
   508             }
   509             return false;
   510           }
   511 
   512           // Augmenting along a shortest path from s to t.
   513           Capacity d = _excess[s] < -_excess[t] ? _excess[s] : -_excess[t];
   514           Node u = t;
   515           Edge e;
   516           if (d > _delta) {
   517             while ((e = _pred[u]) != INVALID) {
   518               Capacity rc;
   519               if (u == _graph.target(e)) {
   520                 rc = _res_cap[e];
   521                 u = _graph.source(e);
   522               } else {
   523                 rc = _flow[e];
   524                 u = _graph.target(e);
   525               }
   526               if (rc < d) d = rc;
   527             }
   528           }
   529           u = t;
   530           while ((e = _pred[u]) != INVALID) {
   531             if (u == _graph.target(e)) {
   532               _flow[e] += d;
   533               _res_cap[e] -= d;
   534               u = _graph.source(e);
   535             } else {
   536               _flow[e] -= d;
   537               _res_cap[e] += d;
   538               u = _graph.target(e);
   539             }
   540           }
   541           _excess[s] -= d;
   542           _excess[t] += d;
   543 
   544           if (_excess[s] < _delta) ++next_node;
   545         }
   546 
   547         if (_delta == 1) break;
   548         if (++phase_cnt > _phase_num / 4) factor = 2;
   549         _delta = _delta <= factor ? 1 : _delta / factor;
   550       }
   551 
   552       // Handling non-zero lower bounds
   553       if (_lower) {
   554         for (EdgeIt e(_graph); e != INVALID; ++e)
   555           _flow[e] += (*_lower)[e];
   556       }
   557       return true;
   558     }
   559 
   560     /// Executes the successive shortest path algorithm.
   561     bool startWithoutScaling() {
   562       // Finding excess nodes
   563       for (NodeIt n(_graph); n != INVALID; ++n)
   564         if (_excess[n] > 0) _excess_nodes.push_back(n);
   565       if (_excess_nodes.size() == 0) return true;
   566       int next_node = 0;
   567 
   568       // Finding shortest paths
   569       Node s, t;
   570       while ( _excess[_excess_nodes[next_node]] > 0 ||
   571               ++next_node < int(_excess_nodes.size()) )
   572       {
   573         // Running Dijkstra
   574         s = _excess_nodes[next_node];
   575         if ((t = _dijkstra.run(s, 1)) == INVALID)
   576           return false;
   577 
   578         // Augmenting along a shortest path from s to t
   579         Capacity d = _excess[s] < -_excess[t] ? _excess[s] : -_excess[t];
   580         Node u = t;
   581         Edge e;
   582         while ((e = _pred[u]) != INVALID) {
   583           Capacity rc;
   584           if (u == _graph.target(e)) {
   585             rc = _res_cap[e];
   586             u = _graph.source(e);
   587           } else {
   588             rc = _flow[e];
   589             u = _graph.target(e);
   590           }
   591           if (rc < d) d = rc;
   592         }
   593         u = t;
   594         while ((e = _pred[u]) != INVALID) {
   595           if (u == _graph.target(e)) {
   596             _flow[e] += d;
   597             _res_cap[e] -= d;
   598             u = _graph.source(e);
   599           } else {
   600             _flow[e] -= d;
   601             _res_cap[e] += d;
   602             u = _graph.target(e);
   603           }
   604         }
   605         _excess[s] -= d;
   606         _excess[t] += d;
   607       }
   608 
   609       // Handling non-zero lower bounds
   610       if (_lower) {
   611         for (EdgeIt e(_graph); e != INVALID; ++e)
   612           _flow[e] += (*_lower)[e];
   613       }
   614       return true;
   615     }
   616 
   617   }; //class CapacityScaling
   618 
   619   ///@}
   620 
   621 } //namespace lemon
   622 
   623 #endif //LEMON_CAPACITY_SCALING_H