org.jgrapht.alg.matching.HopcroftKarpMaximumCardinalityBipartiteMatching Maven / Gradle / Ivy
/*
* (C) Copyright 2017-2021, by Joris Kinable and Contributors.
*
* JGraphT : a free Java graph-theory library
*
* See the CONTRIBUTORS.md file distributed with this work for additional
* information regarding copyright ownership.
*
* This program and the accompanying materials are made available under the
* terms of the Eclipse Public License 2.0 which is available at
* http://www.eclipse.org/legal/epl-2.0, or the
* GNU Lesser General Public License v2.1 or later
* which is available at
* http://www.gnu.org/licenses/old-licenses/lgpl-2.1-standalone.html.
*
* SPDX-License-Identifier: EPL-2.0 OR LGPL-2.1-or-later
*/
package org.jgrapht.alg.matching;
import org.jgrapht.*;
import org.jgrapht.alg.interfaces.*;
import org.jgrapht.alg.util.*;
import java.util.*;
/**
* Implementation of the well-known Hopcroft Karp algorithm to compute a matching of maximum
* cardinality in a bipartite graph. The algorithm runs in $O(|E| \cdot \sqrt{|V|})$ time. This
* implementation accepts undirected graphs which may contain self-loops and multiple edges. To
* compute a maximum cardinality matching in general (non-bipartite) graphs, use
* {@link SparseEdmondsMaximumCardinalityMatching} instead.
*
*
* The Hopcroft Karp matching algorithm computes augmenting paths of increasing length, until no
* augmenting path exists in the graph. At each iteration, the algorithm runs a Breadth First Search
* from the exposed (free) vertices, until an augmenting path is found. Next, a Depth First Search
* is performed to find all (vertex disjoint) augmenting paths of the same length. The matching is
* augmented along all discovered augmenting paths simultaneously.
*
*
* The original algorithm is described in: Hopcroft, John E.; Karp, Richard M. (1973), "An n5/2
* algorithm for maximum matchings in bipartite graphs", SIAM Journal on Computing 2 (4): 225–231,
* doi:10.1137/0202019 A coarse overview of the algorithm is given in: http://en.wikipedia.org/wiki/Hopcroft-Karp_algorithm
*
*
* @param the graph vertex type
* @param the graph edge type
*
* @author Joris Kinable
*/
public class HopcroftKarpMaximumCardinalityBipartiteMatching
implements
MatchingAlgorithm
{
private final Graph graph;
private final Set partition1;
private final Set partition2;
/* Ordered list of vertices */
private List vertices;
/* Mapping of a vertex to their unique position in the ordered list of vertices */
private Map vertexIndexMap;
/* Number of matched vertices i partition 1. */
private int matchedVertices;
/* Dummy vertex. All vertices are initially matched against this dummy vertex */
private static final int DUMMY = 0;
/* Infinity */
private static final int INF = Integer.MAX_VALUE;
/* Array keeping track of the matching. */
private int[] matching;
/* Distance array. Used to compute shoretest augmenting paths */
private int[] dist;
/* queue used for breadth first search */
private FixedSizeIntegerQueue queue;
/**
* Constructs a new instance of the Hopcroft Karp bipartite matching algorithm. The input graph
* must be bipartite. For efficiency reasons, this class does not check whether the input graph
* is bipartite. Invoking this class on a non-bipartite graph results in undefined behavior. To
* test whether a graph is bipartite, use {@link GraphTests#isBipartite(Graph)}.
*
* @param graph bipartite graph
* @param partition1 the first partition of vertices in the bipartite graph
* @param partition2 the second partition of vertices in the bipartite graph
*/
public HopcroftKarpMaximumCardinalityBipartiteMatching(
Graph graph, Set partition1, Set partition2)
{
this.graph = GraphTests.requireUndirected(graph);
// Ensure that partition1 is smaller or equal in size compared to partition 2
if (partition1.size() <= partition2.size()) {
this.partition1 = partition1;
this.partition2 = partition2;
} else { // else, swap
this.partition1 = partition2;
this.partition2 = partition1;
}
}
/**
* Initialize data structures
*/
private void init()
{
vertices = new ArrayList<>();
vertices.add(null);
vertices.addAll(partition1);
vertices.addAll(partition2);
vertexIndexMap = new HashMap<>();
for (int i = 0; i < vertices.size(); i++)
vertexIndexMap.put(vertices.get(i), i);
matching = new int[vertices.size() + 1];
dist = new int[partition1.size() + 1];
queue = new FixedSizeIntegerQueue(vertices.size());
}
/**
* Greedily compute an initial feasible matching
*/
private void warmStart()
{
for (V uOrig : partition1) {
int u = vertexIndexMap.get(uOrig);
for (V vOrig : Graphs.neighborListOf(graph, uOrig)) {
int v = vertexIndexMap.get(vOrig);
if (matching[v] == DUMMY) {
matching[v] = u;
matching[u] = v;
matchedVertices++;
break;
}
}
}
}
/**
* BFS function which finds the shortest augmenting path. The length of the shortest augmenting
* path is stored in dist[DUMMY].
*
* @return true if an augmenting path was found, false otherwise
*/
private boolean bfs()
{
queue.clear();
for (int u = 1; u <= partition1.size(); u++)
if (matching[u] == DUMMY) { // Add all unmatched vertices to the queue and set their
// distance to 0
dist[u] = 0;
queue.enqueue(u);
} else // Set distance of all matched vertices to INF
dist[u] = INF;
dist[DUMMY] = INF;
while (!queue.isEmpty()) {
int u = queue.poll();
if (dist[u] < dist[DUMMY])
for (V vOrig : Graphs.neighborListOf(graph, vertices.get(u))) {
int v = vertexIndexMap.get(vOrig);
if (dist[matching[v]] == INF) {
dist[matching[v]] = dist[u] + 1;
queue.enqueue(matching[v]);
}
}
}
return dist[DUMMY] != INF; // Return true if an augmenting path is found
}
/**
* Find all vertex disjoint augmenting paths of length dist[DUMMY]. To find paths of dist[DUMMY]
* length, we simply follow nodes that are 1 distance increments away from each other.
*
* @param u vertex from which the DFS is started
* @return true if an augmenting path from vertex u was found, false otherwise
*/
private boolean dfs(int u)
{
if (u != DUMMY) {
for (V vOrig : Graphs.neighborListOf(graph, vertices.get(u))) {
int v = vertexIndexMap.get(vOrig);
if (dist[matching[v]] == dist[u] + 1)
if (dfs(matching[v])) {
matching[v] = u;
matching[u] = v;
return true;
}
}
// No augmenting path has been found. Set distance of u to INF to ensure that u isn't
// visited again.
dist[u] = INF;
return false;
}
return true;
}
@Override
public Matching getMatching()
{
this.init();
this.warmStart();
while (matchedVertices < partition1.size() && bfs()) {
// Greedily search for vertex disjoint augmenting paths
for (int v = 1; v <= partition1.size() && matchedVertices < partition1.size(); v++)
if (matching[v] == DUMMY) // v is unmatched
if (dfs(v))
matchedVertices++;
}
assert matchedVertices <= partition1.size();
Set edges = new HashSet<>();
for (int i = 0; i < vertices.size(); i++) {
if (matching[i] != DUMMY) {
edges.add(graph.getEdge(vertices.get(i), vertices.get(matching[i])));
}
}
return new MatchingImpl<>(graph, edges, edges.size());
}
}