ceph/src/boost/boost/graph/bc_clustering.hpp

   1 // Copyright 2004 The Trustees of Indiana University.
   2
   3 // Distributed under the Boost Software License, Version 1.0.
   4 // (See accompanying file LICENSE_1_0.txt or copy at
   5 // http://www.boost.org/LICENSE_1_0.txt)
   6
   7 //  Authors: Douglas Gregor
   8 //           Andrew Lumsdaine
   9 #ifndef BOOST_GRAPH_BETWEENNESS_CENTRALITY_CLUSTERING_HPP
  10 #define BOOST_GRAPH_BETWEENNESS_CENTRALITY_CLUSTERING_HPP
  11
  12 #include <boost/algorithm/minmax_element.hpp>
  13 #include <boost/graph/betweenness_centrality.hpp>
  14 #include <boost/graph/graph_traits.hpp>
  15 #include <boost/graph/graph_utility.hpp>
  16 #include <boost/pending/indirect_cmp.hpp>
  17 #include <vector>
  18 #include <boost/property_map/property_map.hpp>
  19
  20 namespace boost
  21 {
  22
  23 /** Threshold termination function for the betweenness centrality
  24  * clustering algorithm.
  25  */
  26 template < typename T > struct bc_clustering_threshold
  27 {
  28     typedef T centrality_type;
  29
  30     /// Terminate clustering when maximum absolute edge centrality is
  31     /// below the given threshold.
  32     explicit bc_clustering_threshold(T threshold)
  33     : threshold(threshold), dividend(1.0)
  34     {
  35     }
  36
  37     /**
  38      * Terminate clustering when the maximum edge centrality is below
  39      * the given threshold.
  40      *
  41      * @param threshold the threshold value
  42      *
  43      * @param g the graph on which the threshold will be calculated
  44      *
  45      * @param normalize when true, the threshold is compared against the
  46      * normalized edge centrality based on the input graph; otherwise,
  47      * the threshold is compared against the absolute edge centrality.
  48      */
  49     template < typename Graph >
  50     bc_clustering_threshold(T threshold, const Graph& g, bool normalize = true)
  51     : threshold(threshold), dividend(1.0)
  52     {
  53         if (normalize)
  54         {
  55             typename graph_traits< Graph >::vertices_size_type n
  56                 = num_vertices(g);
  57             dividend = T((n - 1) * (n - 2)) / T(2);
  58         }
  59     }
  60
  61     /** Returns true when the given maximum edge centrality (potentially
  62      * normalized) falls below the threshold.
  63      */
  64     template < typename Graph, typename Edge >
  65     bool operator()(T max_centrality, Edge, const Graph&)
  66     {
  67         return (max_centrality / dividend) < threshold;
  68     }
  69
  70 protected:
  71     T threshold;
  72     T dividend;
  73 };
  74
  75 /** Graph clustering based on edge betweenness centrality.
  76  *
  77  * This algorithm implements graph clustering based on edge
  78  * betweenness centrality. It is an iterative algorithm, where in each
  79  * step it compute the edge betweenness centrality (via @ref
  80  * brandes_betweenness_centrality) and removes the edge with the
  81  * maximum betweenness centrality. The @p done function object
  82  * determines when the algorithm terminates (the edge found when the
  83  * algorithm terminates will not be removed).
  84  *
  85  * @param g The graph on which clustering will be performed. The type
  86  * of this parameter (@c MutableGraph) must be a model of the
  87  * VertexListGraph, IncidenceGraph, EdgeListGraph, and Mutable Graph
  88  * concepts.
  89  *
  90  * @param done The function object that indicates termination of the
  91  * algorithm. It must be a ternary function object thats accepts the
  92  * maximum centrality, the descriptor of the edge that will be
  93  * removed, and the graph @p g.
  94  *
  95  * @param edge_centrality (UTIL/OUT) The property map that will store
  96  * the betweenness centrality for each edge. When the algorithm
  97  * terminates, it will contain the edge centralities for the
  98  * graph. The type of this property map must model the
  99  * ReadWritePropertyMap concept. Defaults to an @c
 100  * iterator_property_map whose value type is
 101  * @c Done::centrality_type and using @c get(edge_index, g) for the
 102  * index map.
 103  *
 104  * @param vertex_index (IN) The property map that maps vertices to
 105  * indices in the range @c [0, num_vertices(g)). This type of this
 106  * property map must model the ReadablePropertyMap concept and its
 107  * value type must be an integral type. Defaults to
 108  * @c get(vertex_index, g).
 109  */
 110 template < typename MutableGraph, typename Done, typename EdgeCentralityMap,
 111     typename VertexIndexMap >
 112 void betweenness_centrality_clustering(MutableGraph& g, Done done,
 113     EdgeCentralityMap edge_centrality, VertexIndexMap vertex_index)
 114 {
 115     typedef typename property_traits< EdgeCentralityMap >::value_type
 116         centrality_type;
 117     typedef typename graph_traits< MutableGraph >::edge_iterator edge_iterator;
 118     typedef
 119         typename graph_traits< MutableGraph >::edge_descriptor edge_descriptor;
 120
 121     if (has_no_edges(g))
 122         return;
 123
 124     // Function object that compares the centrality of edges
 125     indirect_cmp< EdgeCentralityMap, std::less< centrality_type > > cmp(
 126         edge_centrality);
 127
 128     bool is_done;
 129     do
 130     {
 131         brandes_betweenness_centrality(g,
 132             edge_centrality_map(edge_centrality)
 133                 .vertex_index_map(vertex_index));
 134         std::pair< edge_iterator, edge_iterator > edges_iters = edges(g);
 135         edge_descriptor e
 136             = *boost::first_max_element(edges_iters.first, edges_iters.second, cmp);
 137         is_done = done(get(edge_centrality, e), e, g);
 138         if (!is_done)
 139             remove_edge(e, g);
 140     } while (!is_done && !has_no_edges(g));
 141 }
 142
 143 /**
 144  * \overload
 145  */
 146 template < typename MutableGraph, typename Done, typename EdgeCentralityMap >
 147 void betweenness_centrality_clustering(
 148     MutableGraph& g, Done done, EdgeCentralityMap edge_centrality)
 149 {
 150     betweenness_centrality_clustering(
 151         g, done, edge_centrality, get(vertex_index, g));
 152 }
 153
 154 /**
 155  * \overload
 156  */
 157 template < typename MutableGraph, typename Done >
 158 void betweenness_centrality_clustering(MutableGraph& g, Done done)
 159 {
 160     typedef typename Done::centrality_type centrality_type;
 161     std::vector< centrality_type > edge_centrality(num_edges(g));
 162     betweenness_centrality_clustering(g, done,
 163         make_iterator_property_map(edge_centrality.begin(), get(edge_index, g)),
 164         get(vertex_index, g));
 165 }
 166
 167 } // end namespace boost
 168
 169 #endif // BOOST_GRAPH_BETWEENNESS_CENTRALITY_CLUSTERING_HPP