Back to home page

EIC code displayed by LXR

 
 

    


File indexing completed on 2025-01-30 09:43:00

0001 // Copyright (C) 2006-2009 Dmitry Bufistov and Andrey Parfenov
0002 
0003 // Use, modification and distribution is subject to the Boost Software
0004 // License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
0005 // http://www.boost.org/LICENSE_1_0.txt)
0006 
0007 #ifndef BOOST_GRAPH_CYCLE_RATIO_HOWARD_HPP
0008 #define BOOST_GRAPH_CYCLE_RATIO_HOWARD_HPP
0009 
0010 #include <vector>
0011 #include <list>
0012 #include <algorithm>
0013 #include <functional>
0014 #include <limits>
0015 
0016 #include <boost/bind/bind.hpp>
0017 #include <boost/tuple/tuple.hpp>
0018 #include <boost/type_traits/is_same.hpp>
0019 #include <boost/type_traits/remove_const.hpp>
0020 #include <boost/concept_check.hpp>
0021 #include <boost/pending/queue.hpp>
0022 #include <boost/property_map/property_map.hpp>
0023 #include <boost/graph/graph_traits.hpp>
0024 #include <boost/graph/graph_concepts.hpp>
0025 #include <boost/concept/assert.hpp>
0026 #include <boost/algorithm/minmax_element.hpp>
0027 
0028 /** @file howard_cycle_ratio.hpp
0029  * @brief The implementation of the maximum/minimum cycle ratio/mean algorithm.
0030  * @author Dmitry Bufistov
0031  * @author Andrey Parfenov
0032  */
0033 
0034 namespace boost
0035 {
0036 
0037 /**
0038  * The mcr_float is like numeric_limits, but only for floating point types
0039  * and only defines infinity() and epsilon(). This class is primarily used
0040  * to encapsulate a less-precise epsilon than natively supported by the
0041  * floating point type.
0042  */
0043 template < typename Float = double > struct mcr_float
0044 {
0045     typedef Float value_type;
0046 
0047     static Float infinity()
0048     {
0049         return std::numeric_limits< value_type >::infinity();
0050     }
0051 
0052     static Float epsilon() { return Float(-0.005); }
0053 };
0054 
0055 namespace detail
0056 {
0057 
0058     template < typename FloatTraits > struct min_comparator_props
0059     {
0060         typedef std::greater< typename FloatTraits::value_type > comparator;
0061         static const int multiplier = 1;
0062     };
0063 
0064     template < typename FloatTraits > struct max_comparator_props
0065     {
0066         typedef std::less< typename FloatTraits::value_type > comparator;
0067         static const int multiplier = -1;
0068     };
0069 
0070     template < typename FloatTraits, typename ComparatorProps >
0071     struct float_wrapper
0072     {
0073         typedef typename FloatTraits::value_type value_type;
0074         typedef ComparatorProps comparator_props_t;
0075         typedef typename ComparatorProps::comparator comparator;
0076 
0077         static value_type infinity()
0078         {
0079             return FloatTraits::infinity() * ComparatorProps::multiplier;
0080         }
0081 
0082         static value_type epsilon()
0083         {
0084             return FloatTraits::epsilon() * ComparatorProps::multiplier;
0085         }
0086     };
0087 
0088     /*! @class mcr_howard
0089      * @brief Calculates optimum (maximum/minimum) cycle ratio of a directed
0090      * graph. Uses  Howard's iteration policy algorithm. </br>(It is described
0091      * in the paper "Experimental Analysis of the Fastest Optimum Cycle Ratio
0092      * and Mean Algorithm" by Ali Dasdan).
0093      */
0094     template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0095         typename EdgeWeight1, typename EdgeWeight2 >
0096     class mcr_howard
0097     {
0098     public:
0099         typedef typename FloatTraits::value_type float_t;
0100         typedef typename FloatTraits::comparator_props_t cmp_props_t;
0101         typedef typename FloatTraits::comparator comparator_t;
0102         typedef enum
0103         {
0104             my_white = 0,
0105             my_black
0106         } my_color_type;
0107         typedef typename graph_traits< Graph >::vertex_descriptor vertex_t;
0108         typedef typename graph_traits< Graph >::edge_descriptor edge_t;
0109         typedef typename graph_traits< Graph >::vertices_size_type vn_t;
0110         typedef std::vector< float_t > vp_t;
0111         typedef typename boost::iterator_property_map< typename vp_t::iterator,
0112             VertexIndexMap >
0113             distance_map_t; // V -> float_t
0114 
0115         typedef typename std::vector< edge_t > ve_t;
0116         typedef std::vector< my_color_type > vcol_t;
0117         typedef
0118             typename ::boost::iterator_property_map< typename ve_t::iterator,
0119                 VertexIndexMap >
0120                 policy_t; // Vertex -> Edge
0121         typedef
0122             typename ::boost::iterator_property_map< typename vcol_t::iterator,
0123                 VertexIndexMap >
0124                 color_map_t;
0125 
0126         typedef typename std::list< vertex_t >
0127             pinel_t; // The in_edges list of the policy graph
0128         typedef typename std::vector< pinel_t > inedges1_t;
0129         typedef typename ::boost::iterator_property_map<
0130             typename inedges1_t::iterator, VertexIndexMap >
0131             inedges_t;
0132         typedef typename std::vector< edge_t > critical_cycle_t;
0133 
0134         // Bad  vertex flag. If true, then the vertex is "bad".
0135         // Vertex is "bad" if its out_degree is equal to zero.
0136         typedef
0137             typename boost::iterator_property_map< std::vector< int >::iterator,
0138                 VertexIndexMap >
0139                 badv_t;
0140 
0141         /*!
0142          * Constructor
0143          * \param g = (V, E) - a directed multigraph.
0144          * \param vim  Vertex Index Map. Read property Map: V -> [0,
0145          * num_vertices(g)). \param ewm  edge weight map. Read property map: E
0146          * -> R \param ew2m  edge weight map. Read property map: E -> R+ \param
0147          * infty A big enough value to guaranty that there exist a cycle with
0148          *  better ratio.
0149          * \param cmp The compare operator for float_ts.
0150          */
0151         mcr_howard(const Graph& g, VertexIndexMap vim, EdgeWeight1 ewm,
0152             EdgeWeight2 ew2m)
0153         : m_g(g)
0154         , m_vim(vim)
0155         , m_ew1m(ewm)
0156         , m_ew2m(ew2m)
0157         , m_bound(mcr_bound())
0158         , m_cr(m_bound)
0159         , m_V(num_vertices(m_g))
0160         , m_dis(m_V, 0)
0161         , m_dm(m_dis.begin(), m_vim)
0162         , m_policyc(m_V)
0163         , m_policy(m_policyc.begin(), m_vim)
0164         , m_inelc(m_V)
0165         , m_inel(m_inelc.begin(), m_vim)
0166         , m_badvc(m_V, false)
0167         , m_badv(m_badvc.begin(), m_vim)
0168         , m_colcv(m_V)
0169         , m_col_bfs(m_V)
0170         {
0171         }
0172 
0173         /*!
0174          * \return maximum/minimum_{for all cycles C}
0175          *         [sum_{e in C} w1(e)] / [sum_{e in C} w2(e)],
0176          * or FloatTraits::infinity() if graph has no cycles.
0177          */
0178         float_t ocr_howard()
0179         {
0180             construct_policy_graph();
0181             int k = 0;
0182             float_t mcr = 0;
0183             do
0184             {
0185                 mcr = policy_mcr();
0186                 ++k;
0187             } while (
0188                 try_improve_policy(mcr) && k < 100); // To avoid infinite loop
0189 
0190             const float_t eps_ = -0.00000001 * cmp_props_t::multiplier;
0191             if (m_cmp(mcr, m_bound + eps_))
0192             {
0193                 return FloatTraits::infinity();
0194             }
0195             else
0196             {
0197                 return mcr;
0198             }
0199         }
0200         virtual ~mcr_howard() {}
0201 
0202     protected:
0203         virtual void store_critical_edge(edge_t, critical_cycle_t&) {}
0204         virtual void store_critical_cycle(critical_cycle_t&) {}
0205 
0206     private:
0207         /*!
0208          * \return lower/upper bound for the maximal/minimal cycle ratio
0209          */
0210         float_t mcr_bound()
0211         {
0212             typename graph_traits< Graph >::vertex_iterator vi, vie;
0213             typename graph_traits< Graph >::out_edge_iterator oei, oeie;
0214             float_t cz = (std::numeric_limits< float_t >::max)(); // Closest to
0215                                                                   // zero value
0216             float_t s = 0;
0217             const float_t eps_ = std::numeric_limits< float_t >::epsilon();
0218             for (boost::tie(vi, vie) = vertices(m_g); vi != vie; ++vi)
0219             {
0220                 for (boost::tie(oei, oeie) = out_edges(*vi, m_g); oei != oeie;
0221                      ++oei)
0222                 {
0223                     s += std::abs(m_ew1m[*oei]);
0224                     float_t a = std::abs(m_ew2m[*oei]);
0225                     if (a > eps_ && a < cz)
0226                     {
0227                         cz = a;
0228                     }
0229                 }
0230             }
0231             return cmp_props_t::multiplier * (s / cz);
0232         }
0233 
0234         /*!
0235          *  Constructs an arbitrary policy graph.
0236          */
0237         void construct_policy_graph()
0238         {
0239             m_sink = graph_traits< Graph >().null_vertex();
0240             typename graph_traits< Graph >::vertex_iterator vi, vie;
0241             typename graph_traits< Graph >::out_edge_iterator oei, oeie;
0242             for (boost::tie(vi, vie) = vertices(m_g); vi != vie; ++vi)
0243             {
0244                 using namespace boost::placeholders;
0245 
0246                 boost::tie(oei, oeie) = out_edges(*vi, m_g);
0247                 typename graph_traits< Graph >::out_edge_iterator mei
0248                     = boost::first_max_element(oei, oeie,
0249                         boost::bind(m_cmp,
0250                             boost::bind(&EdgeWeight1::operator[], m_ew1m, _1),
0251                             boost::bind(&EdgeWeight1::operator[], m_ew1m, _2)));
0252                 if (mei == oeie)
0253                 {
0254                     if (m_sink == graph_traits< Graph >().null_vertex())
0255                     {
0256                         m_sink = *vi;
0257                     }
0258                     m_badv[*vi] = true;
0259                     m_inel[m_sink].push_back(*vi);
0260                 }
0261                 else
0262                 {
0263                     m_inel[target(*mei, m_g)].push_back(*vi);
0264                     m_policy[*vi] = *mei;
0265                 }
0266             }
0267         }
0268         /*! Sets the distance value for all vertices "v" such that there is
0269          * a path from "v" to "sv". It does "inverse" breadth first visit of the
0270          * policy graph, starting from the vertex "sv".
0271          */
0272         void mcr_bfv(vertex_t sv, float_t cr, color_map_t c)
0273         {
0274             boost::queue< vertex_t > Q;
0275             c[sv] = my_black;
0276             Q.push(sv);
0277             while (!Q.empty())
0278             {
0279                 vertex_t v = Q.top();
0280                 Q.pop();
0281                 for (typename pinel_t::const_iterator itr = m_inel[v].begin();
0282                      itr != m_inel[v].end(); ++itr)
0283                 // For all in_edges of the policy graph
0284                 {
0285                     if (*itr != sv)
0286                     {
0287                         if (m_badv[*itr])
0288                         {
0289                             m_dm[*itr] = m_dm[v] + m_bound - cr;
0290                         }
0291                         else
0292                         {
0293                             m_dm[*itr] = m_dm[v] + m_ew1m[m_policy[*itr]]
0294                                 - m_ew2m[m_policy[*itr]] * cr;
0295                         }
0296                         c[*itr] = my_black;
0297                         Q.push(*itr);
0298                     }
0299                 }
0300             }
0301         }
0302 
0303         /*!
0304          * \param sv an arbitrary (undiscovered) vertex of the policy graph.
0305          * \return a vertex in the policy graph that belongs to a cycle.
0306          * Performs a depth first visit until a cycle edge is found.
0307          */
0308         vertex_t find_cycle_vertex(vertex_t sv)
0309         {
0310             vertex_t gv = sv;
0311             std::fill(m_colcv.begin(), m_colcv.end(), my_white);
0312             color_map_t cm(m_colcv.begin(), m_vim);
0313             do
0314             {
0315                 cm[gv] = my_black;
0316                 if (!m_badv[gv])
0317                 {
0318                     gv = target(m_policy[gv], m_g);
0319                 }
0320                 else
0321                 {
0322                     gv = m_sink;
0323                 }
0324             } while (cm[gv] != my_black);
0325             return gv;
0326         }
0327 
0328         /*!
0329          * \param sv - vertex that belongs to a cycle in the policy graph.
0330          */
0331         float_t cycle_ratio(vertex_t sv)
0332         {
0333             if (sv == m_sink)
0334                 return m_bound;
0335             std::pair< float_t, float_t > sums_(float_t(0), float_t(0));
0336             vertex_t v = sv;
0337             critical_cycle_t cc;
0338             do
0339             {
0340                 store_critical_edge(m_policy[v], cc);
0341                 sums_.first += m_ew1m[m_policy[v]];
0342                 sums_.second += m_ew2m[m_policy[v]];
0343                 v = target(m_policy[v], m_g);
0344             } while (v != sv);
0345             float_t cr = sums_.first / sums_.second;
0346             if (m_cmp(m_cr, cr))
0347             {
0348                 m_cr = cr;
0349                 store_critical_cycle(cc);
0350             }
0351             return cr;
0352         }
0353 
0354         /*!
0355          *  Finds the optimal cycle ratio of the policy graph
0356          */
0357         float_t policy_mcr()
0358         {
0359             using namespace boost::placeholders;
0360 
0361             std::fill(m_col_bfs.begin(), m_col_bfs.end(), my_white);
0362             color_map_t vcm_ = color_map_t(m_col_bfs.begin(), m_vim);
0363             typename graph_traits< Graph >::vertex_iterator uv_itr, vie;
0364             boost::tie(uv_itr, vie) = vertices(m_g);
0365             float_t mcr = m_bound;
0366             while ((uv_itr = std::find_if(uv_itr, vie,
0367                         boost::bind(std::equal_to< my_color_type >(), my_white,
0368                             boost::bind(&color_map_t::operator[], vcm_, _1))))
0369                 != vie)
0370             /// While there are undiscovered vertices
0371             {
0372                 vertex_t gv = find_cycle_vertex(*uv_itr);
0373                 float_t cr = cycle_ratio(gv);
0374                 mcr_bfv(gv, cr, vcm_);
0375                 if (m_cmp(mcr, cr))
0376                     mcr = cr;
0377                 ++uv_itr;
0378             }
0379             return mcr;
0380         }
0381 
0382         /*!
0383          * Changes the edge m_policy[s] to the new_edge.
0384          */
0385         void improve_policy(vertex_t s, edge_t new_edge)
0386         {
0387             vertex_t t = target(m_policy[s], m_g);
0388             typename property_traits< VertexIndexMap >::value_type ti
0389                 = m_vim[t];
0390             m_inelc[ti].erase(
0391                 std::find(m_inelc[ti].begin(), m_inelc[ti].end(), s));
0392             m_policy[s] = new_edge;
0393             t = target(new_edge, m_g);
0394             m_inel[t].push_back(s); /// Maintain in_edge list
0395         }
0396 
0397         /*!
0398          * A negative cycle detector.
0399          */
0400         bool try_improve_policy(float_t cr)
0401         {
0402             bool improved = false;
0403             typename graph_traits< Graph >::vertex_iterator vi, vie;
0404             typename graph_traits< Graph >::out_edge_iterator oei, oeie;
0405             const float_t eps_ = FloatTraits::epsilon();
0406             for (boost::tie(vi, vie) = vertices(m_g); vi != vie; ++vi)
0407             {
0408                 if (!m_badv[*vi])
0409                 {
0410                     for (boost::tie(oei, oeie) = out_edges(*vi, m_g);
0411                          oei != oeie; ++oei)
0412                     {
0413                         vertex_t t = target(*oei, m_g);
0414                         // Current distance from *vi to some vertex
0415                         float_t dis_
0416                             = m_ew1m[*oei] - m_ew2m[*oei] * cr + m_dm[t];
0417                         if (m_cmp(m_dm[*vi] + eps_, dis_))
0418                         {
0419                             improve_policy(*vi, *oei);
0420                             m_dm[*vi] = dis_;
0421                             improved = true;
0422                         }
0423                     }
0424                 }
0425                 else
0426                 {
0427                     float_t dis_ = m_bound - cr + m_dm[m_sink];
0428                     if (m_cmp(m_dm[*vi] + eps_, dis_))
0429                     {
0430                         m_dm[*vi] = dis_;
0431                     }
0432                 }
0433             }
0434             return improved;
0435         }
0436 
0437     private:
0438         const Graph& m_g;
0439         VertexIndexMap m_vim;
0440         EdgeWeight1 m_ew1m;
0441         EdgeWeight2 m_ew2m;
0442         comparator_t m_cmp;
0443         float_t m_bound; //> The lower/upper bound to the maximal/minimal cycle
0444                          // ratio
0445         float_t m_cr; //>The best cycle ratio that has been found so far
0446 
0447         vn_t m_V; //>The number of the vertices in the graph
0448         vp_t m_dis; //>Container for the distance map
0449         distance_map_t m_dm; //>Distance map
0450 
0451         ve_t m_policyc; //>Container for the policy graph
0452         policy_t m_policy; //>The interface for the policy graph
0453 
0454         inedges1_t m_inelc; //>Container fot in edges list
0455         inedges_t m_inel; //>Policy graph, input edges list
0456 
0457         std::vector< int > m_badvc;
0458         badv_t m_badv; // Marks "bad" vertices
0459 
0460         vcol_t m_colcv, m_col_bfs; // Color maps
0461         vertex_t m_sink; // To convert any graph to "good"
0462     };
0463 
0464     /*! \class mcr_howard1
0465      * \brief Finds optimum cycle raio and a critical cycle
0466      */
0467     template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0468         typename EdgeWeight1, typename EdgeWeight2 >
0469     class mcr_howard1 : public mcr_howard< FloatTraits, Graph, VertexIndexMap,
0470                             EdgeWeight1, EdgeWeight2 >
0471     {
0472     public:
0473         typedef mcr_howard< FloatTraits, Graph, VertexIndexMap, EdgeWeight1,
0474             EdgeWeight2 >
0475             inhr_t;
0476         mcr_howard1(const Graph& g, VertexIndexMap vim, EdgeWeight1 ewm,
0477             EdgeWeight2 ew2m)
0478         : inhr_t(g, vim, ewm, ew2m)
0479         {
0480         }
0481 
0482         void get_critical_cycle(typename inhr_t::critical_cycle_t& cc)
0483         {
0484             return cc.swap(m_cc);
0485         }
0486 
0487     protected:
0488         void store_critical_edge(
0489             typename inhr_t::edge_t ed, typename inhr_t::critical_cycle_t& cc)
0490         {
0491             cc.push_back(ed);
0492         }
0493 
0494         void store_critical_cycle(typename inhr_t::critical_cycle_t& cc)
0495         {
0496             m_cc.swap(cc);
0497         }
0498 
0499     private:
0500         typename inhr_t::critical_cycle_t m_cc; // Critical cycle
0501     };
0502 
0503     /*!
0504      * \param g a directed multigraph.
0505      * \param vim Vertex Index Map. A map V->[0, num_vertices(g))
0506      * \param ewm Edge weight1 map.
0507      * \param ew2m Edge weight2 map.
0508      * \param pcc  pointer to the critical edges list.
0509      * \return Optimum cycle ratio of g or FloatTraits::infinity() if g has no
0510      * cycles.
0511      */
0512     template < typename FT, typename TG, typename TVIM, typename TEW1,
0513         typename TEW2, typename EV >
0514     typename FT::value_type optimum_cycle_ratio(
0515         const TG& g, TVIM vim, TEW1 ewm, TEW2 ew2m, EV* pcc)
0516     {
0517         typedef typename graph_traits< TG >::directed_category DirCat;
0518         BOOST_STATIC_ASSERT(
0519             (is_convertible< DirCat*, directed_tag* >::value == true));
0520         BOOST_CONCEPT_ASSERT((IncidenceGraphConcept< TG >));
0521         BOOST_CONCEPT_ASSERT((VertexListGraphConcept< TG >));
0522         typedef typename graph_traits< TG >::vertex_descriptor Vertex;
0523         BOOST_CONCEPT_ASSERT((ReadablePropertyMapConcept< TVIM, Vertex >));
0524         typedef typename graph_traits< TG >::edge_descriptor Edge;
0525         BOOST_CONCEPT_ASSERT((ReadablePropertyMapConcept< TEW1, Edge >));
0526         BOOST_CONCEPT_ASSERT((ReadablePropertyMapConcept< TEW2, Edge >));
0527 
0528         if (pcc == 0)
0529         {
0530             return detail::mcr_howard< FT, TG, TVIM, TEW1, TEW2 >(
0531                 g, vim, ewm, ew2m)
0532                 .ocr_howard();
0533         }
0534 
0535         detail::mcr_howard1< FT, TG, TVIM, TEW1, TEW2 > obj(g, vim, ewm, ew2m);
0536         double ocr = obj.ocr_howard();
0537         obj.get_critical_cycle(*pcc);
0538         return ocr;
0539     }
0540 } // namespace detail
0541 
0542 // Algorithms
0543 // Maximum Cycle Ratio
0544 
0545 template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0546     typename EdgeWeight1Map, typename EdgeWeight2Map >
0547 inline typename FloatTraits::value_type maximum_cycle_ratio(const Graph& g,
0548     VertexIndexMap vim, EdgeWeight1Map ew1m, EdgeWeight2Map ew2m,
0549     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0,
0550     FloatTraits = FloatTraits())
0551 {
0552     typedef detail::float_wrapper< FloatTraits,
0553         detail::max_comparator_props< FloatTraits > >
0554         Traits;
0555     return detail::optimum_cycle_ratio< Traits >(g, vim, ew1m, ew2m, pcc);
0556 }
0557 
0558 template < typename Graph, typename VertexIndexMap, typename EdgeWeight1Map,
0559     typename EdgeWeight2Map >
0560 inline double maximum_cycle_ratio(const Graph& g, VertexIndexMap vim,
0561     EdgeWeight1Map ew1m, EdgeWeight2Map ew2m,
0562     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0)
0563 {
0564     return maximum_cycle_ratio(g, vim, ew1m, ew2m, pcc, mcr_float<>());
0565 }
0566 
0567 // Minimum Cycle Ratio
0568 
0569 template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0570     typename EdgeWeight1Map, typename EdgeWeight2Map >
0571 typename FloatTraits::value_type minimum_cycle_ratio(const Graph& g,
0572     VertexIndexMap vim, EdgeWeight1Map ew1m, EdgeWeight2Map ew2m,
0573     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0,
0574     FloatTraits = FloatTraits())
0575 {
0576     typedef detail::float_wrapper< FloatTraits,
0577         detail::min_comparator_props< FloatTraits > >
0578         Traits;
0579     return detail::optimum_cycle_ratio< Traits >(g, vim, ew1m, ew2m, pcc);
0580 }
0581 
0582 template < typename Graph, typename VertexIndexMap, typename EdgeWeight1Map,
0583     typename EdgeWeight2Map >
0584 inline double minimum_cycle_ratio(const Graph& g, VertexIndexMap vim,
0585     EdgeWeight1Map ew1m, EdgeWeight2Map ew2m,
0586     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0)
0587 {
0588     return minimum_cycle_ratio(g, vim, ew1m, ew2m, pcc, mcr_float<>());
0589 }
0590 
0591 // Maximum Cycle Mean
0592 
0593 template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0594     typename EdgeWeightMap, typename EdgeIndexMap >
0595 inline typename FloatTraits::value_type maximum_cycle_mean(const Graph& g,
0596     VertexIndexMap vim, EdgeWeightMap ewm, EdgeIndexMap eim,
0597     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0,
0598     FloatTraits ft = FloatTraits())
0599 {
0600     typedef typename remove_const<
0601         typename property_traits< EdgeWeightMap >::value_type >::type Weight;
0602     typename std::vector< Weight > ed_w2(boost::num_edges(g), 1);
0603     return maximum_cycle_ratio(
0604         g, vim, ewm, make_iterator_property_map(ed_w2.begin(), eim), pcc, ft);
0605 }
0606 
0607 template < typename Graph, typename VertexIndexMap, typename EdgeWeightMap,
0608     typename EdgeIndexMap >
0609 inline double maximum_cycle_mean(const Graph& g, VertexIndexMap vim,
0610     EdgeWeightMap ewm, EdgeIndexMap eim,
0611     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0)
0612 {
0613     return maximum_cycle_mean(g, vim, ewm, eim, pcc, mcr_float<>());
0614 }
0615 
0616 // Minimum Cycle Mean
0617 
0618 template < typename FloatTraits, typename Graph, typename VertexIndexMap,
0619     typename EdgeWeightMap, typename EdgeIndexMap >
0620 inline typename FloatTraits::value_type minimum_cycle_mean(const Graph& g,
0621     VertexIndexMap vim, EdgeWeightMap ewm, EdgeIndexMap eim,
0622     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0,
0623     FloatTraits ft = FloatTraits())
0624 {
0625     typedef typename remove_const<
0626         typename property_traits< EdgeWeightMap >::value_type >::type Weight;
0627     typename std::vector< Weight > ed_w2(boost::num_edges(g), 1);
0628     return minimum_cycle_ratio(
0629         g, vim, ewm, make_iterator_property_map(ed_w2.begin(), eim), pcc, ft);
0630 }
0631 
0632 template < typename Graph, typename VertexIndexMap, typename EdgeWeightMap,
0633     typename EdgeIndexMap >
0634 inline double minimum_cycle_mean(const Graph& g, VertexIndexMap vim,
0635     EdgeWeightMap ewm, EdgeIndexMap eim,
0636     std::vector< typename graph_traits< Graph >::edge_descriptor >* pcc = 0)
0637 {
0638     return minimum_cycle_mean(g, vim, ewm, eim, pcc, mcr_float<>());
0639 }
0640 
0641 } // namespace boost
0642 
0643 #endif