首页 > 代码库 > HDU2686-Matrix & HDU3376-Matrix Again(费用流)
HDU2686-Matrix & HDU3376-Matrix Again(费用流)
比较简单的题了。
只需从左上角到右下角找两条路就可以了。
因为每个点只能走一次,所以拆点,限制流量为1。
因为求的是最大值,所以权值取反求最小值。
因为第一个点和最后一个点经过两次,只算一次,最后要减去。
ps:数组还是开大点好。。。。不知道什么时候就SB了。。。
注意汇点可能不是最后一个点(模板的问题。。
注意初始化的范围。。。。
matrix again只是数据大了,算法不用改。。。无聊。。。。
#include <algorithm>#include <iostream>#include <cstring>#include <string>#include <vector>#include <bitset>#include <cstdio>#include <queue>#include <stack>#include <cmath>#include <list>#include <map>#include <set>#define pk(x) printf("%d\n", x)using namespace std;#define PI acos(-1.0)#define EPS 1E-6#define clr(x,c) memset(x,c,sizeof(x))//#pragma comment(linker, "/STACK:102400000,102400000")typedef long long ll;const int MAXV = 2000;const int INF = 1<<30;struct Edge { int to, cap, cost, rev; };vector<Edge> G[MAXV];int dist[MAXV], prv[MAXV], pre[MAXV], in[MAXV];queue<int> que;void addedge(int from, int to, int cap, int cost) { G[from].push_back((Edge){to, cap, cost, G[to].size()}); G[to].push_back((Edge){from, 0, -cost, G[from].size()-1});}int min_cost_max_flow(int s, int t) { //, int f) { int res = 0; int f = 0; while (1) { //f > 0) { for (int i = 1; i <= t; ++i) dist[i] = INF, in[i] = 0; dist[s] = 0; while (!que.empty()) que.pop(); in[s] = 1; que.push(s); while (!que.empty()) { int u = que.front(); que.pop(); in[u] = 0; for (int i = 0; i < G[u].size(); ++i) { Edge &e = G[u][i]; if (e.cap > 0 && dist[e.to] > dist[u] + e.cost) { dist[e.to] = dist[u] + e.cost; prv[e.to] = u; pre[e.to] = i; if (in[e.to] == 0) { in[e.to] = 1; que.push(e.to); } } } } if (dist[t] == INF) break; //return -1; int d = INF; // d = f; for (int v = t; v != s; v = prv[v]) { d = min(d, G[prv[v]][pre[v]].cap); } f += d; res += d * dist[t]; for (int v = t; v != s; v = prv[v]) { Edge &e = G[prv[v]][pre[v]]; e.cap -= d; G[v][e.rev].cap += d; } } return res; //return f;}int n;int getid1(int x, int y) { return x*n+y+1; }int getid2(int x, int y) { return x*n+y+n*n+1; }int a[30][30];int main(){ while (~scanf("%d", &n)) { for (int i = 0; i < n; ++i) { for (int j = 0; j < n; ++j) { scanf("%d", &a[i][j]); } } for (int i = 0; i <= 2*n*n; ++i) G[i].clear(); for (int i = 0; i < n; ++i) { for (int j = 0; j < n; ++j) { if (i+j == 0) addedge(getid1(i,j), getid2(i,j), 2, -a[i][j]); else if (i+j == n*2 - 2) addedge(getid1(i,j), getid2(i, j), 2, -a[i][j]); else addedge(getid1(i, j), getid2(i, j), 1, -a[i][j]); if (i+1<n) addedge(getid2(i, j), getid1(i+1, j), 1, 0); if (j+1<n) addedge(getid2(i, j), getid1(i, j+1), 1, 0); } } printf("%d\n", -min_cost_max_flow(getid1(0,0), getid2(n-1,n-1)) - a[0][0] - a[n-1][n-1]); } return 0;}
#include <algorithm>#include <iostream>#include <cstring>#include <string>#include <vector>#include <bitset>#include <cstdio>#include <queue>#include <stack>#include <cmath>#include <list>#include <map>#include <set>#define pk(x) printf("%d\n", x)using namespace std;#define PI acos(-1.0)#define EPS 1E-6#define clr(x,c) memset(x,c,sizeof(x))//#pragma comment(linker, "/STACK:102400000,102400000")typedef long long ll;const int MAXV = 610*610*2+2;const int INF = 1<<30;struct Edge { int to, cap, cost, rev; };vector<Edge> G[MAXV];int dist[MAXV], prv[MAXV], pre[MAXV], in[MAXV];queue<int> que;void addedge(int from, int to, int cap, int cost) { G[from].push_back((Edge){to, cap, cost, G[to].size()}); G[to].push_back((Edge){from, 0, -cost, G[from].size()-1});}int min_cost_max_flow(int s, int t) { //, int f) { int res = 0; int f = 0; while (1) { //f > 0) { for (int i = 1; i <= t; ++i) dist[i] = INF, in[i] = 0; dist[s] = 0; while (!que.empty()) que.pop(); in[s] = 1; que.push(s); while (!que.empty()) { int u = que.front(); que.pop(); in[u] = 0; for (int i = 0; i < G[u].size(); ++i) { Edge &e = G[u][i]; if (e.cap > 0 && dist[e.to] > dist[u] + e.cost) { dist[e.to] = dist[u] + e.cost; prv[e.to] = u; pre[e.to] = i; if (in[e.to] == 0) { in[e.to] = 1; que.push(e.to); } } } } if (dist[t] == INF) break; //return -1; int d = INF; // d = f; for (int v = t; v != s; v = prv[v]) { d = min(d, G[prv[v]][pre[v]].cap); } f += d; res += d * dist[t]; for (int v = t; v != s; v = prv[v]) { Edge &e = G[prv[v]][pre[v]]; e.cap -= d; G[v][e.rev].cap += d; } } return res; //return f;}inline int Scan(){ char ch = getchar(); int data = http://www.mamicode.com/0; while (ch < ‘0‘ || ch > ‘9‘) ch = getchar(); do { data = data*10 + ch-‘0‘; ch = getchar(); } while (ch >= ‘0‘ && ch <= ‘9‘); return data;}int n;int getid1(int x, int y) { return x*n+y+1; }int getid2(int x, int y) { return x*n+y+n*n+1; }int a[600][600];int main(){ while (~scanf("%d", &n)) { int maxn = 2*n*n; for (int i = 0; i <= maxn; ++i) G[i].clear(); for (int i = 0; i < n; ++i) { for (int j = 0; j < n; ++j) { a[i][j] = Scan(); if (i+j == 0 || i+j == n*2-2) addedge(getid1(i,j), getid2(i,j), 2, -a[i][j]); else addedge(getid1(i, j), getid2(i, j), 1, -a[i][j]); if (i+1<n) addedge(getid2(i, j), getid1(i+1, j), 1, 0); if (j+1<n) addedge(getid2(i, j), getid1(i, j+1), 1, 0); } } printf("%d\n", -min_cost_max_flow(getid1(0,0), getid2(n-1,n-1)) - a[0][0] - a[n-1][n-1]); } return 0;}
HDU2686-Matrix & HDU3376-Matrix Again(费用流)
声明:以上内容来自用户投稿及互联网公开渠道收集整理发布,本网站不拥有所有权,未作人工编辑处理,也不承担相关法律责任,若内容有误或涉及侵权可进行投诉: 投诉/举报 工作人员会在5个工作日内联系你,一经查实,本站将立刻删除涉嫌侵权内容。