Algorithmic Solutions > LEDA > LEDA Guide > Graph Algorithms > Matching Algorithms > Bipartite Weighted Matching > Example MAX_WEIGHT_BIPARTITE_MATCHING()

Example MAX_WEIGHT_BIPARTITE_MATCHING()

The following program shows how the function MAX_WEIGHT_BIPARTITE_MATCHING() can be used to compute a maximum weighted matching and a corresponding potential function for the nodes in a bipartite graph. It also shows how the function CHECK_MCB() can be used to check the correctness of the result.

Remark: The graph algorithms in LEDA are generic, that is, they accept graphs as well as parameterized graphs.

In main() we first create a bipartite graph G=(A,B,E) with three nodes in A, four nodes in B, and six edges. We use a list<node> to store A and B. We use an edge_array<double> weight to store the weights of the edges of G. . The variant of MAX_WEIGHT_BIPARTITE_MATCHING() for int can be used in exactly the same way. You only need to replace double by int in weight and pot.

#include <LEDA/graph/graph.h>
#include <LEDA/graph/mwb_matching.h>

using namespace leda;

int main()
{
  graph G; 

  list<node> B;
  node n0=G.new_node(); B.append(n0);
  node n1=G.new_node(); B.append(n1);
  node n2=G.new_node(); B.append(n2);
  node n3=G.new_node(); B.append(n3);

  list<node> A;
  node n4=G.new_node(); A.append(n4);
  node n5=G.new_node(); A.append(n5);
  node n6=G.new_node(); A.append(n6);

  edge e0=G.new_edge(n4,n0); edge e1=G.new_edge(n4,n1);
  edge e2=G.new_edge(n4,n2); edge e3=G.new_edge(n5,n2);
  edge e4=G.new_edge(n6,n2); edge e5=G.new_edge(n6,n3);
 
  edge_array<double> weight(G);
  weight[e0]=1; weight[e1]=1; weight[e2]=3;
  weight[e3]=1; weight[e4]=1; weight[e5]=1;

In order to avoid that MAX_WEIGHT_BIPARTITE_MATCHING() modifies the edge weights internally, we call MWBM_SCALE_WEIGHTS() explicitely. In this small example the weights are, of course, unchanged and MWBM_SCALE_WEIGHTS() returns true. We only want to demonstrate the usage here.

The result of MAX_WEIGHT_BIPARTITE_MATCHING() is a list<edge> M containing the edges in the maximum weighted matching and a node_array<integer> pot for the potentials of the nodes of G.

We use CHECK_MCB_T() to check if M is a maximum weighted matching of G and pot is a corresponding potential function. If the result is correct we output M and pot.

  bool unmodified_weights=MWBM_SCALE_WEIGHTS(G,weight);

  if (!unmodified_weights) {
	  std::cout << "Warning: MWBM_SCALE_WEIGHTS()"
	            << " modified your edge weights!" << std::endl;
  }

  node_array<double> pot(G);
  list<edge> M;
  M=MAX_WEIGHT_BIPARTITE_MATCHING(G,A,B,weight,pot);

  bool result_correct=CHECK_MWBM(G,weight,M,pot);

  if (result_correct) {
    std::cout << "Maximum Weighted Matching:" << std::endl;
    double weight_of_M=0;
    edge e;
    forall(e,M) {G.print_edge(e); weight_of_M+=weight[e];}
    std::cout << " weight: " << weight_of_M << std::endl;

    node v; forall_nodes(v,G) {
      cstd::out << "pot"; G.print_node(v);
      std::cout << "=" << pot[v] << std::endl;
	}
  }

  else std::cout << "Error: M and pot are not"
                 << " a correct solution!" << std::endl;

  return 0;
}

Remark: There are variants of MAX_WEIGHT_MATCHING() that do not need the parameter pot and variants without explicitely stating the bipartition of G as list<node> A, list<node> B. Have look at the Manual Page Bipartite Weighted Matchings and Assignments for details.

Tip: Using the smaller set as A and the bigger set as B leads to smaller running times, in general.

See also:

Bipartite Weighted Matching

Graphs

Parameterized Graphs

Linear Lists

Edge Arrays

Node Arrays


Matching Algorithms


Manual Entries:

Manual Page Bipartite Weighted Matchings and Assignments



Please send any suggestions, comments or questions to leda@algorithmic-solutions.com
© Copyright 2001-2003, Algorithmic Solutions Software GmbH