Венгерский алгоритм - C#
Формулировка задачи:
Ребята прошу Вашей помощи, нужен исходник Венгерского алгоритма на C#
Решение задачи: «Венгерский алгоритм»
textual
Листинг программы
- using System;
- public class OptimalAssignmentSolver {
- /// <summary>
- /// Solves optimal assignment problem using the Kuhn-Munkres
- /// (aka Hungarian) algorithm. Time complexity: O(n^3).
- ///
- /// References:
- /// [url]http://www.math.uwo.ca/~mdawes/courses/344/kuhn-munkres.pdf[/url]
- /// [url]http://www.ecp6.jussieu.fr/pageperso/bondy/books/gtwa/pdf/chapter5.pdf[/url]
- /// </summary>
- /// <param name="a">A square matrix.</param>
- /// <returns>
- /// A permutation p of integers 0, 1, ..., n-1, (where
- /// n is the size of matrix a) such that the expression
- /// a[0, p[0]] + a[1, p[1]] + ... + a[n-1, p[n-1]]
- /// is maximum possible among all such permutations.
- /// </returns>
- public static int[] KuhnMunkres(int[,] a) {
- int N = a.GetLength(0);
- if (N == 0)
- return new int[0];
- int[] lx = new int[N], ly = new int[N]; // labelling function for vertices in first and second partitions
- int[] mx = new int[N], my = new int[N]; // mx[u]=v, my[v]=u <==> u and v are currently matched; -1 values means 'not matched'
- int[] px = new int[N], py = new int[N]; // predecessor arrays. used in DFS to reconstruct paths.
- int[] stack = new int[N];
- // invariant: lx[u] + ly[v] >= a[u, v]
- // (implies that any perfect matching in subgraph containing only
- // edges u, v for which lx[u]+ly[v]=a[u,v] is the optimal matching.)
- // compute initial labelling function: lx[i] = max_j(a[i, j]), ly[j] = 0;
- for (int i = 0; i < N; i++) {
- lx[i] = a[i, 0];
- for (int j = 0; j < N; j++)
- if (a[i, j] > lx[i]) lx[i] = a[i, j];
- ly[i] = 0;
- mx[i] = my[i] = -1;
- }
- for (int size = 0; size < N;) {
- int s;
- for (s = 0; mx[s] != -1; s++);
- // s is an unmatched vertex in the first partition.
- // At the current iteration we will either find an augmenting path
- // starting at s, or we'll extend the equality subgraph so that
- // such a path will exist at the next iteration.
- for (int i = 0; i < N; i++)
- px[i] = py[i] = -1;
- px[s] = s;
- // DFS
- int t = -1;
- stack[0] = s;
- for (int top = 1; top > 0;) {
- int u = stack[--top];
- for (int v = 0; v < N; v++) {
- if (lx[u] + ly[v] == a[u, v] && py[v] == -1) {
- if (my[v] == -1) {
- // we've found an augmenting path
- t = v;
- py[t] = u;
- top = 0;
- break;
- }
- py[v] = u;
- px[my[v]] = v;
- stack[top++] = my[v];
- }
- }
- }
- if (t != -1) {
- // augment along the found path
- while (true) {
- int u = py[t];
- mx[u] = t;
- my[t] = u;
- if (u == s) break;
- t = px[u];
- }
- ++size;
- } else {
- // No augmenting path exists from s in the current equality graph,
- // Modify labelling function a bit...
- int delta = int.MaxValue;
- for (int u = 0; u < N; u++) {
- if (px[u] == -1) continue;
- for (int v = 0; v < N; v++) {
- if (py[v] != -1) continue;
- int z = lx[u] + ly[v] - a[u, v];
- if (z < delta)
- delta = z;
- }
- }
- for (int i = 0; i < N; i++) {
- if (px[i] != -1) lx[i] -= delta;
- if (py[i] != -1) ly[i] += delta;
- }
- }
- }
- // Verify optimality
- bool correct = true;
- for (int u = 0; u < N; u++) {
- for (int v = 0; v < N; v++) {
- correct &= (lx[u] + ly[v] >= a[u, v]);
- if (mx[u] == v)
- correct &= (lx[u] + ly[v] == a[u, v]);
- if (!correct) {
- throw new Exception(
- "*** Internal error: optimality conditions are not satisfied ***\n" +
- "Most probably an overflow occurred");
- }
- }
- }
- return mx;
- }
- }
ИИ поможет Вам:
- решить любую задачу по программированию
- объяснить код
- расставить комментарии в коде
- и т.д