lemon/gomory_hu_tree.h
author Janos Tapolcai <tapolcai@tmit.bme.hu>
Fri, 20 Feb 2009 17:17:17 +0100
changeset 590 924887566bf2
child 591 ccd2d3a3001e
permissions -rw-r--r--
Porting Gomory-Hu algorithm (#66)
     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_GOMORY_HU_TREE_H
    20 #define LEMON_GOMORY_HU_TREE_H
    21 
    22 #include <limits>
    23 
    24 #include <lemon/preflow.h>
    25 #include <lemon/concept_check.h>
    26 #include <lemon/concepts/maps.h>
    27 
    28 /// \ingroup min_cut
    29 /// \file 
    30 /// \brief Gomory-Hu cut tree in graphs.
    31 
    32 namespace lemon {
    33 
    34   /// \ingroup min_cut
    35   ///
    36   /// \brief Gomory-Hu cut tree algorithm
    37   ///
    38   /// The Gomory-Hu tree is a tree on the nodeset of the digraph, but it
    39   /// may contain arcs which are not in the original digraph. It helps
    40   /// to calculate the minimum cut between all pairs of nodes, because
    41   /// the minimum capacity arc on the tree path between two nodes has
    42   /// the same weight as the minimum cut in the digraph between these
    43   /// nodes. Moreover this arc separates the nodes to two parts which
    44   /// determine this minimum cut.
    45   /// 
    46   /// The algorithm calculates \e n-1 distinict minimum cuts with
    47   /// preflow algorithm, therefore the algorithm has
    48   /// \f$(O(n^3\sqrt{e})\f$ overall time complexity. It calculates a
    49   /// rooted Gomory-Hu tree, the structure of the tree and the weights
    50   /// can be obtained with \c predNode() and \c predValue()
    51   /// functions. The \c minCutValue() and \c minCutMap() calculates
    52   /// the minimum cut and the minimum cut value between any two node
    53   /// in the digraph.
    54   template <typename _Graph, 
    55 	    typename _Capacity = typename _Graph::template EdgeMap<int> >
    56   class GomoryHuTree {
    57   public:
    58 
    59     /// The graph type
    60     typedef _Graph Graph;
    61     /// The capacity on edges
    62     typedef _Capacity Capacity;
    63     /// The value type of capacities
    64     typedef typename Capacity::Value Value;
    65     
    66   private:
    67 
    68     TEMPLATE_GRAPH_TYPEDEFS(Graph);
    69 
    70     const Graph& _graph;
    71     const Capacity& _capacity;
    72 
    73     Node _root;
    74     typename Graph::template NodeMap<Node>* _pred;
    75     typename Graph::template NodeMap<Value>* _weight;
    76     typename Graph::template NodeMap<int>* _order;
    77 
    78     void createStructures() {
    79       if (!_pred) {
    80 	_pred = new typename Graph::template NodeMap<Node>(_graph);
    81       }
    82       if (!_weight) {
    83 	_weight = new typename Graph::template NodeMap<Value>(_graph);
    84       }
    85       if (!_order) {
    86 	_order = new typename Graph::template NodeMap<int>(_graph);
    87       }
    88     }
    89 
    90     void destroyStructures() {
    91       if (_pred) {
    92 	delete _pred;
    93       }
    94       if (_weight) {
    95 	delete _weight;
    96       }
    97       if (_order) {
    98 	delete _order;
    99       }
   100     }
   101   
   102   public:
   103 
   104     /// \brief Constructor
   105     ///
   106     /// Constructor
   107     /// \param graph The graph type.
   108     /// \param capacity The capacity map.
   109     GomoryHuTree(const Graph& graph, const Capacity& capacity) 
   110       : _graph(graph), _capacity(capacity),
   111 	_pred(0), _weight(0), _order(0) 
   112     {
   113       checkConcept<concepts::ReadMap<Edge, Value>, Capacity>();
   114     }
   115 
   116 
   117     /// \brief Destructor
   118     ///
   119     /// Destructor
   120     ~GomoryHuTree() {
   121       destroyStructures();
   122     }
   123 
   124     /// \brief Initializes the internal data structures.
   125     ///
   126     /// Initializes the internal data structures.
   127     ///
   128     void init() {
   129       createStructures();
   130 
   131       _root = NodeIt(_graph);
   132       for (NodeIt n(_graph); n != INVALID; ++n) {
   133 	_pred->set(n, _root);
   134 	_order->set(n, -1);
   135       }
   136       _pred->set(_root, INVALID);
   137       _weight->set(_root, std::numeric_limits<Value>::max()); 
   138     }
   139 
   140 
   141     /// \brief Starts the algorithm
   142     ///
   143     /// Starts the algorithm.
   144     void start() {
   145       Preflow<Graph, Capacity> fa(_graph, _capacity, _root, INVALID);
   146 
   147       for (NodeIt n(_graph); n != INVALID; ++n) {
   148 	if (n == _root) continue;
   149 
   150 	Node pn = (*_pred)[n];
   151 	fa.source(n);
   152 	fa.target(pn);
   153 
   154 	fa.runMinCut();
   155 
   156 	_weight->set(n, fa.flowValue());
   157 
   158 	for (NodeIt nn(_graph); nn != INVALID; ++nn) {
   159 	  if (nn != n && fa.minCut(nn) && (*_pred)[nn] == pn) {
   160 	    _pred->set(nn, n);
   161 	  }
   162 	}
   163 	if ((*_pred)[pn] != INVALID && fa.minCut((*_pred)[pn])) {
   164 	  _pred->set(n, (*_pred)[pn]);
   165 	  _pred->set(pn, n);
   166 	  _weight->set(n, (*_weight)[pn]);
   167 	  _weight->set(pn, fa.flowValue());	
   168 	}
   169       }
   170 
   171       _order->set(_root, 0);
   172       int index = 1;
   173 
   174       for (NodeIt n(_graph); n != INVALID; ++n) {
   175 	std::vector<Node> st;
   176 	Node nn = n;
   177 	while ((*_order)[nn] == -1) {
   178 	  st.push_back(nn);
   179 	  nn = (*_pred)[nn];
   180 	}
   181 	while (!st.empty()) {
   182 	  _order->set(st.back(), index++);
   183 	  st.pop_back();
   184 	}
   185       }
   186     }
   187 
   188     /// \brief Runs the Gomory-Hu algorithm.  
   189     ///
   190     /// Runs the Gomory-Hu algorithm.
   191     /// \note gh.run() is just a shortcut of the following code.
   192     /// \code
   193     ///   ght.init();
   194     ///   ght.start();
   195     /// \endcode
   196     void run() {
   197       init();
   198       start();
   199     }
   200 
   201     /// \brief Returns the predecessor node in the Gomory-Hu tree.
   202     ///
   203     /// Returns the predecessor node in the Gomory-Hu tree. If the node is
   204     /// the root of the Gomory-Hu tree, then it returns \c INVALID.
   205     Node predNode(const Node& node) {
   206       return (*_pred)[node];
   207     }
   208 
   209     /// \brief Returns the weight of the predecessor arc in the
   210     /// Gomory-Hu tree.
   211     ///
   212     /// Returns the weight of the predecessor arc in the Gomory-Hu
   213     /// tree.  If the node is the root of the Gomory-Hu tree, the
   214     /// result is undefined.
   215     Value predValue(const Node& node) {
   216       return (*_weight)[node];
   217     }
   218 
   219     /// \brief Returns the minimum cut value between two nodes
   220     ///
   221     /// Returns the minimum cut value between two nodes. The
   222     /// algorithm finds the nearest common ancestor in the Gomory-Hu
   223     /// tree and calculates the minimum weight arc on the paths to
   224     /// the ancestor.
   225     Value minCutValue(const Node& s, const Node& t) const {
   226       Node sn = s, tn = t;
   227       Value value = std::numeric_limits<Value>::max();
   228       
   229       while (sn != tn) {
   230 	if ((*_order)[sn] < (*_order)[tn]) {
   231 	  if ((*_weight)[tn] < value) value = (*_weight)[tn];
   232 	  tn = (*_pred)[tn];
   233 	} else {
   234 	  if ((*_weight)[sn] < value) value = (*_weight)[sn];
   235 	  sn = (*_pred)[sn];
   236 	}
   237       }
   238       return value;
   239     }
   240 
   241     /// \brief Returns the minimum cut between two nodes
   242     ///
   243     /// Returns the minimum cut value between two nodes. The
   244     /// algorithm finds the nearest common ancestor in the Gomory-Hu
   245     /// tree and calculates the minimum weight arc on the paths to
   246     /// the ancestor. Then it sets all nodes to the cut determined by
   247     /// this arc. The \c cutMap should be \ref concepts::ReadWriteMap
   248     /// "ReadWriteMap".
   249     template <typename CutMap>
   250     Value minCutMap(const Node& s, const Node& t, CutMap& cutMap) const {
   251       Node sn = s, tn = t;
   252 
   253       Node rn = INVALID;
   254       Value value = std::numeric_limits<Value>::max();
   255       
   256       while (sn != tn) {
   257 	if ((*_order)[sn] < (*_order)[tn]) {
   258 	  if ((*_weight)[tn] < value) {
   259 	    rn = tn;
   260 	    value = (*_weight)[tn];
   261 	  }
   262 	  tn = (*_pred)[tn];
   263 	} else {
   264 	  if ((*_weight)[sn] < value) {
   265 	    rn = sn;
   266 	    value = (*_weight)[sn];
   267 	  }
   268 	  sn = (*_pred)[sn];
   269 	}
   270       }
   271 
   272       typename Graph::template NodeMap<bool> reached(_graph, false);
   273       reached.set(_root, true);
   274       cutMap.set(_root, false);
   275       reached.set(rn, true);
   276       cutMap.set(rn, true);
   277 
   278       for (NodeIt n(_graph); n != INVALID; ++n) {
   279 	std::vector<Node> st;
   280 	Node nn = n;
   281 	while (!reached[nn]) {
   282 	  st.push_back(nn);
   283 	  nn = (*_pred)[nn];
   284 	}
   285 	while (!st.empty()) {
   286 	  cutMap.set(st.back(), cutMap[nn]);
   287 	  st.pop_back();
   288 	}
   289       }
   290       
   291       return value;
   292     }
   293 
   294   };
   295 
   296 }
   297 
   298 #endif