lemon-0.x: src/work/athos/preflow

     1 #ifndef PREFLOW_PUSH_HH

     2 #define PREFLOW_PUSH_HH

     4 #include <algorithm>

     5 #include <list>

     6 #include <vector>

     7 //#include "pf_hiba.hh"

     8 //#include <marci_list_graph.hh>

     9 //#include <marci_graph_traits.hh>

    11 #include "reverse_bfs.hh"

    13 using namespace std;

    15 namespace marci {

    17   template <typename graph_type, typename T>

    18   class preflow_push {

    20     //Hasznos typedef-ek

    21     typedef typename graph_type::NodeIt NodeIt;

    22     typedef typename graph_type::EdgeIt EdgeIt;

    23     typedef typename graph_type::EachNodeIt EachNodeIt;

    24     typedef typename graph_type::EachEdgeIt EachEdgeIt;

    25     typedef typename graph_type::OutEdgeIt OutEdgeIt;

    26     typedef typename graph_type::InEdgeIt InEdgeIt;

    27     typedef typename graph_type::SymEdgeIt SymEdgeIt;

    31     /*

    32     typedef graph_traits<graph_type>::node_iterator node_iterator;

    33     typedef graph_traits<graph_type>::EdgeIt EdgeIt;

    34     typedef graph_traits<graph_type>::each_node_iterator each_node_iterator;

    35     typedef graph_traits<graph_type>::each_EdgeIt each_EdgeIt;

    36     typedef graph_traits<graph_type>::out_EdgeIt out_EdgeIt;

    37     typedef graph_traits<graph_type>::InEdgeIt InEdgeIt;

    38     typedef graph_traits<graph_type>::sym_EdgeIt sym_EdgeIt;

    39     */

    41     //---------------------------------------------

    42     //Parameters of the algorithm

    43     //---------------------------------------------

    44     //Fully examine an active node until excess becomes 0

    45     enum node_examination_t {examine_full, examine_to_relabel};

    46     //No more implemented yet:, examine_only_one_edge};

    47     node_examination_t node_examination;

    48     //Which implementation to be used

    49     enum implementation_t {impl_fifo, impl_highest_label};

    50     //No more implemented yet:};

    51     implementation_t implementation;

    52     //---------------------------------------------

    53     //Parameters of the algorithm

    54     //---------------------------------------------

    56   private:

    57     //input

    58     graph_type& G;

    59     NodeIt s;

    60     NodeIt t;

    61     typename graph_type::EdgeMap<T> &capacity;

    62     //typename graph_type::EdgeMap<T>  &capacity;

    63     //output

    64     //typename graph_type::EdgeMap<T>

    65     typename graph_type::EdgeMap<T> preflow;

    66     T maxflow_value;

    68     //auxiliary variables for computation

    69     int number_of_nodes;

    72     typename graph_type::NodeMap<int> level;

    73     typename graph_type::NodeMap<T> excess;

    74     //node_property_vector<graph_type, int> level;

    75     //node_property_vector<graph_type, T> excess;

    77     //Number of nodes on each level

    78     vector<int> num_of_nodes_on_level;

    80     //For the FIFO implementation

    81     list<NodeIt> fifo_nodes;

    82     //For 'highest label' implementation

    83     int highest_active;

    84     //int second_highest_active;

    85     vector< list<NodeIt> > active_nodes;

    87   public:

    89     //Constructing the object using the graph, source, sink and capacity vector

    90     preflow_push(

    91 		      graph_type& _G,

    92 		      NodeIt _s,

    93 		      NodeIt _t,

    94 		      typename graph_type::EdgeMap<T> & _capacity)

    95       : G(_G), s(_s), t(_t),

    96 	capacity(_capacity),

    97 	preflow(_G),

    98 	//Counting the number of nodes

    99 	//number_of_nodes(count(G.first<EachNodeIt>())),

   100 	number_of_nodes(G.nodeNum()),

   102 	level(_G),

   103 	excess(_G)//,

   104         // Default constructor: active_nodes()

   105     {

   106       //Simplest parameter settings

   107       node_examination = examine_full;//examine_to_relabel;//

   108       //Which implementation to be usedexamine_full

   109       implementation = impl_highest_label;//impl_fifo;

   111       //

   112       num_of_nodes_on_level.resize(2*number_of_nodes-1);

   113       num_of_nodes_on_level.clear();

   115       switch(implementation){

   116       case impl_highest_label :{

   117 	active_nodes.resize(2*number_of_nodes-1);

   118 	active_nodes.clear();

   119 	break;

   120       }

   121       default:

   122 	break;

   123       }

   125     }

   127     //Returns the value of a maximal flow

   128     T run();

   130     typename graph_type::EdgeMap<T>  getmaxflow(){

   131       return preflow;

   132     }

   135   private:

   136     //For testing purposes only

   137     //Lists the node_properties

   138     void write_property_vector(typename graph_type::NodeMap<T> a,

   139 			       //node_property_vector<graph_type, T> a,

   140 			       char* prop_name="property"){

   141       for(EachNodeIt i=G.template first<EachNodeIt>(); i.valid(); ++i) {

   142 	cout<<"Node id.: "<<G.id(i)<<", "<<prop_name<<" value: "<<a.get(i)<<endl;

   143       }

   144       cout<<endl;

   145     }

   147     //Modifies the excess of the node and makes sufficient changes

   148     void modify_excess(const NodeIt& a ,T v){

   149 	T old_value=excess.get(a);

   150 	excess.set(a,old_value+v);

   151     }

   153     //This private procedure is supposed to modify the preflow on edge j

   154     //by value v (which can be positive or negative as well)

   155     //and maintain the excess on the head and tail

   156     //Here we do not check whether this is possible or not

   157     void modify_preflow(EdgeIt j, const T& v){

   159       //Auxiliary variable

   160       T old_value;

   162       //Modifiyng the edge

   163       old_value=preflow.get(j);

   164       preflow.set(j,old_value+v);

   167       //Modifiyng the head

   168       modify_excess(G.head(j),v);

   170       //Modifiyng the tail

   171       modify_excess(G.tail(j),-v);

   173     }

   175     //Gives the active node to work with

   176     //(depending on the implementation to be used)

   177     NodeIt get_active_node(){

   178       //cout<<highest_active<<endl;

   180       switch(implementation) {

   181       case impl_highest_label : {

   183 	//First need to find the highest label for which there"s an active node

   184 	while( highest_active>=0 && active_nodes[highest_active].empty() ){

   185 	  --highest_active;

   186 	}

   188 	if( highest_active>=0) {

   189 	  NodeIt a=active_nodes[highest_active].front();

   190 	  active_nodes[highest_active].pop_front();

   191 	  return a;

   192 	}

   193 	else {

   194 	  return NodeIt();

   195 	}

   197 	break;

   199       }

   200       case impl_fifo : {

   202 	if( ! fifo_nodes.empty() ) {

   203 	  NodeIt a=fifo_nodes.front();

   204 	  fifo_nodes.pop_front();

   205 	  return a;

   206 	}

   207 	else {

   208 	  return NodeIt();

   209 	}

   210 	break;

   211       }

   212       }

   213       //

   214       return NodeIt();

   215     }

   217     //Puts node 'a' among the active nodes

   218     void make_active(const NodeIt& a){

   219       //s and t never become active

   220       if (a!=s && a!= t){

   221 	switch(implementation){

   222 	case impl_highest_label :

   223 	  active_nodes[level.get(a)].push_back(a);

   224 	  break;

   225 	case impl_fifo :

   226 	  fifo_nodes.push_back(a);

   227 	  break;

   228 	}

   230       }

   232       //Update highest_active label

   233       if (highest_active<level.get(a)){

   234 	highest_active=level.get(a);

   235       }

   237     }

   239     //Changes the level of node a and make sufficent changes

   240     void change_level_to(NodeIt a, int new_value){

   241       int seged = level.get(a);

   242       level.set(a,new_value);

   243       --num_of_nodes_on_level[seged];

   244       ++num_of_nodes_on_level[new_value];

   245     }

   247     //Collection of things useful (or necessary) to do before running

   249     void preprocess(){

   251       //---------------------------------------

   252       //Initialize parameters

   253       //---------------------------------------

   255       //Setting starting preflow, level and excess values to zero

   256       //This can be important, if the algorithm is run more then once

   257       for(EachNodeIt i=G.template first<EachNodeIt>(); i.valid(); ++i) {

   258         level.set(i,0);

   259         excess.set(i,0);

   260 	for(OutEdgeIt j=G.template first<OutEdgeIt>(i); j.valid(); ++j)

   261 	  preflow.set(j, 0);

   262       }

   263       num_of_nodes_on_level[0]=number_of_nodes;

   264       highest_active=0;

   265       //---------------------------------------

   266       //Initialize parameters

   267       //---------------------------------------

   270       //------------------------------------

   271       //This is the only part that uses BFS

   272       //------------------------------------

   273       //Setting starting level values using reverse bfs

   274       reverse_bfs<graph_type> rev_bfs(G,t);

   275       rev_bfs.run();

   276       //write_property_vector(rev_bfs.dist,"rev_bfs");

   277       for(EachNodeIt i=G.template first<EachNodeIt>(); i.valid(); ++i) {

   278         change_level_to(i,rev_bfs.dist(i));

   279 	//level.put(i,rev_bfs.dist.get(i));

   280       }

   281       //------------------------------------

   282       //This is the only part that uses BFS

   283       //------------------------------------

   286       //Starting level of s

   287       change_level_to(s,number_of_nodes);

   288       //level.put(s,number_of_nodes);

   291       //we push as much preflow from s as possible to start with

   292       for(OutEdgeIt j=G.template first<OutEdgeIt>(s); j.valid(); ++j){

   293 	modify_preflow(j,capacity.get(j) );

   294 	make_active(G.head(j));

   295 	int lev=level.get(G.head(j));

   296 	if(highest_active<lev){

   297 	  highest_active=lev;

   298 	}

   299       }

   300       //cout<<highest_active<<endl;

   301     }

   304     //If the preflow is less than the capacity on the given edge

   305     //then it is an edge in the residual graph

   306     bool is_admissible_forward_edge(OutEdgeIt j, int& new_level){

   307       if (capacity.get(j)>preflow.get(j)){

   308 	if(level.get(G.tail(j))==level.get(G.head(j))+1){

   309 	  return true;

   310 	}

   311 	else{

   312 	  if (level.get(G.head(j)) < new_level)

   313 	    new_level=level.get(G.head(j));

   314 	}

   315       }

   316       return false;

   317     }

   319     //If the preflow is greater than 0 on the given edge

   320     //then the edge reversd is an edge in the residual graph

   321     bool is_admissible_backward_edge(InEdgeIt j, int& new_level){

   322       if (0<preflow.get(j)){

   323 	if(level.get(G.tail(j))==level.get(G.head(j))-1){

   324 	  return true;

   325 	}

   326 	else{

   327 	  if (level.get(G.tail(j)) < new_level)

   328 	    new_level=level.get(G.tail(j));

   329 	}

   331       }

   332       return false;

   333     }

   336   };  //class preflow_push

   338   template<typename graph_type, typename T>

   339     T preflow_push<graph_type, T>::run() {

   341     preprocess();

   343     T e,v;

   344     NodeIt a;

   345     while (a=get_active_node(), a.valid()){

   346       //cout<<G.id(a)<<endl;

   347       //write_property_vector(excess,"excess");

   348       //write_property_vector(level,"level");

   351       bool go_to_next_node=false;

   352       e = excess.get(a);

   353       while (!go_to_next_node){

   355 	//Initial value for the new level for the active node we are dealing with

   356 	int new_level=2*number_of_nodes;

   357 	//write_property_vector(excess,"excess");

   358 	//write_property_vector(level,"level");

   359 	//cout<<G.id(a)<<endl;

   360 	//Out edges from node a

   361 	{

   362 	  OutEdgeIt j=G.template first<OutEdgeIt>(a);

   363 	  while (j.valid() && e){

   365 	    if (is_admissible_forward_edge(j,new_level)){

   366 	      v=min(e,capacity.get(j) - preflow.get(j));

   367 	      e -= v;

   368 	      //New node might become active

   369 	      if (excess.get(G.head(j))==0){

   370 		make_active(G.head(j));

   371 	      }

   372 	      modify_preflow(j,v);

   373 	    }

   374 	    ++j;

   375 	  }

   376 	}

   377 	//In edges to node a

   378 	{

   379 	  InEdgeIt j=G.template first<InEdgeIt>(a);

   380 	  while (j.valid() && e){

   381 	    if (is_admissible_backward_edge(j,new_level)){

   382 	      v=min(e,preflow.get(j));

   383 	      e -= v;

   384 	      //New node might become active

   385 	      if (excess.get(G.tail(j))==0){

   386 		make_active(G.tail(j));

   387 	      }

   388 	      modify_preflow(j,-v);

   389 	    }

   390 	    ++j;

   391 	  }

   392 	}

   394 	//cout<<G.id(a)<<" "<<new_level<<endl;

   396 	if (0==e){

   397 	  //Saturating push

   398 	  go_to_next_node=true;

   399 	}

   400 	else{//If there is still excess in node a

   402 	  //change_level_to(a,new_level+1);

   404 	  //Level remains empty

   405 	  if (num_of_nodes_on_level[level.get(a)]==1){

   406 	    change_level_to(a,number_of_nodes);

   407 	    //go_to_next_node=True;

   408 	  }

   409 	  else{

   410 	    change_level_to(a,new_level+1);

   411 	    //increase_level(a);

   412 	  }

   417 	  switch(node_examination){

   418 	  case examine_to_relabel:

   419 	    make_active(a);

   421 	    go_to_next_node = true;

   422 	    break;

   423 	  default:

   424 	    break;

   425 	  }

   429 	}//if (0==e)

   430       }

   431     }

   432     maxflow_value = excess.get(t);

   433     return maxflow_value;

   434   }//run

   437 }//namespace marci

   439 #endif //PREFLOW_PUSH_HH

author	marci
	Wed, 18 Feb 2004 12:45:44 +0000
changeset 95	3322fbf254d2
parent 36	7d539ea6ad26
child 105	a3c73e9b9b2e
permissions	-rw-r--r--