Network Wars


Time Limit: 5 Seconds      Memory Limit: 32768 KB      Special Judge

Network of Byteland consists of n servers, connected by m optical cables. Each cable connects two servers and can transmit data in both directions. Two servers of the network are especially important --- they are connected to global world network and president palace network respectively.

The server connected to the president palace network has number 1, and the server connected to the global world network has number n.

Recently the company Max Traffic has decided to take control over some cables so that it could see what data is transmitted by the president palace users. Of course they want to control such set of cables, that it is impossible to download any data from the global network to the president palace without transmitting it over at least one of the cables from the set.

To put its plans into practice the company needs to buy corresponding cables from their current owners. Each cable has some cost. Since the company's main business is not spying, but providing internet connection to home users, its management wants to make the operation a good investment. So it wants to buy such a set of cables, that cables mean cost} is minimal possible.

That is, if the company buys k cables of the total cost c, it wants to minimize the value of c/k.

Input

There are several test cases in the input. The first line of each case contains n and m (2 <= n <= 100 , 1 <= m <= 400 ). Next m lines describe cables~--- each cable is described with three integer numbers: servers it connects and the cost of the cable. Cost of each cable is positive and does not exceed107.

Any two servers are connected by at most one cable. No cable connects a server to itself. The network is guaranteed to be connected, it is possible to transmit data from any server to any other one.

There is an empty line between each cases.

Output

First output k --- the number of cables to buy. After that output the cables to buy themselves. Cables are numbered starting from one in order they are given in the input file. There should an empty line between each cases.

Example

Input Output
6 8
1 2 3
1 3 3
2 4 2
2 5 2
3 4 2
3 5 2
5 6 3
4 6 3
4
3 4 5 6
4 5
1 2 2
1 3 2
2 3 1
2 4 2
3 4 2
3
1 2 3

题目链接:ZOJ 2676

此题叫我们求$\Sigma w_{ei} \over |E|$的最小值,其中所有的边均在S-T的割中,可以发现当${\Sigma w_{ei} \over |E|}<r$时,存在$r'={\Sigma w_{ei} \over |E|}$作为更优的r,那我们写成$\Sigma w_{ei} - r*|E|<0$,存在一个左边的结果使得等式成立,即找到左边式子的最小值小于0即可,观察左边的式子,可以化简成$\Sigma (w_{ei}-r)<0$,然后边集e是一个割,又要求这个割集的最小值,那显然就是求s-t的最小割即可,先用二分求出最佳的比例,然后在最后剩下的那个残余网络中找出割集。

代码:

#include <bits/stdc++.h>
using namespace std;
#define INF 0x3f3f3f3f
#define LC(x) (x<<1)
#define RC(x) ((x<<1)+1)
#define MID(x,y) ((x+y)>>1)
#define CLR(arr,val) memset(arr,val,sizeof(arr))
#define FAST_IO ios::sync_with_stdio(false);cin.tie(0);
typedef pair<int, int> pii;
typedef long long LL;
const double PI = acos(-1.0);
const int N = 110;
const int M = 410;
const double eps = 1e-6;
struct edge
{
int to, nxt;
double cap;
edge() {}
edge(int _to, int _nxt, double _cap): to(_to), nxt(_nxt), cap(_cap) {}
};
struct Node
{
int u, v;
double cap;
};
Node e[M];
edge E[M << 1];
int head[N], tot;
int d[N];
int use[M]; void init()
{
CLR(head, -1);
tot = 0;
CLR(use, 0);
}
inline void add(int s, int t, double cap)
{
E[tot] = edge(t, head[s], cap);
head[s] = tot++;
E[tot] = edge(s, head[t], cap);
head[t] = tot++;
}
int bfs(int s, int t)
{
CLR(d, -1);
d[s] = 0;
queue<int>Q;
Q.push(s);
while (!Q.empty())
{
int u = Q.front();
Q.pop();
for (int i = head[u]; ~i; i = E[i].nxt)
{
int v = E[i].to;
if (d[v] == -1 && E[i].cap > 0)
{
d[v] = d[u] + 1;
if (v == t)
return 1;
Q.push(v);
}
}
}
return ~d[t];
}
double dfs(int s, int t, double f)
{
if (s == t || !f)
return f;
double ret = 0;
for (int i = head[s]; ~i; i = E[i].nxt)
{
int v = E[i].to;
if (d[v] == d[s] + 1 && E[i].cap > 0)
{
double df = dfs(v, t, min(f, E[i].cap));
if (df > 0)
{
E[i].cap -= df;
E[i ^ 1].cap += df;
f -= df;
ret += df;
if (!f)
break;
}
}
}
if (!ret)
d[s] = -1;
return ret;
}
double dinic(int s, int t)
{
double ans = 0;
while (bfs(s, t))
ans += dfs(s, t, INF);
return ans;
}
double Mincut(int n, int m, double r)
{
int i;
init();
double ret = 0;
for (i = 1; i <= m; ++i)
{
if (e[i].cap < r)
{
ret += e[i].cap - r;
use[i] = 1;
}
else
add(e[i].u, e[i].v, e[i].cap - r);
}
return ret + dinic(1, n);
}
int main(void)
{
int n, m, i;
while (~scanf("%d%d", &n, &m))
{
for (i = 1; i <= m; ++i)
scanf("%d%d%lf", &e[i].u, &e[i].v, &e[i].cap);
double Rat = 0, L = 0, R = 400.0 / 3 * 1e7;
while (fabs(R - L) >= eps)
{
double mid = (L + R) / 2.0;
if (Mincut(n, m, mid) < 0)
{
R = mid;
Rat = mid;
}
else
L = mid;
}
vector<int>ans;
for (i = 1; i <= m; ++i)
{
if ((d[e[i].u]!=-1)^(d[e[i].v]!=-1))
use[i] = 1;
if (use[i])
ans.push_back(i);
}
int sz = ans.size();
printf("%d\n", sz);
for (i = 0; i < sz; ++i)
printf("%d%c", ans[i], " \n"[i == sz - 1]);
}
return 0;
}

ZOJ 2676 Network Wars(最优比例最小割)的更多相关文章

  1. zoj 2676 Network Wars 0-1分数规划+最小割

    题目详解出自 论文 Amber-最小割模型在信息学竞赛中的应用 题目大意: 给出一个带权无向图 G = (V,E), 每条边 e属于E都有一个权值We,求一个割边集C,使得该割边集的平均边权最小,即最 ...

  2. HDU 2676 Network Wars 01分数规划,最小割 难度:4

    http://acm.zju.edu.cn/onlinejudge/showProblem.do?problemId=1676 对顶点i,j,起点s=1,终点t=n,可以认为题意要求一组01矩阵use ...

  3. ZOJ 2676 Network Wars[01分数规划]

    ZOJ Problem Set - 2676 Network Wars Time Limit: 5 Seconds      Memory Limit: 32768 KB      Special J ...

  4. ZOJ 2676 Network Wars ★(最小割算法介绍 && 01分数规划)

    [题意]给出一个带权无向图,求割集,且割集的平均边权最小. [分析] 先尝试着用更一般的形式重新叙述本问题.设向量w表示边的权值,令向量c=(1, 1, 1, --, 1)表示选边的代价,于是原问题等 ...

  5. ZOJ 2676 Network Wars(网络流+分数规划)

    传送门 题意:求无向图割集中平均边权最小的集合. 论文<最小割模型在信息学竞赛中的应用>原题. 分数规划.每一条边取上的代价为1. #include <bits/stdc++.h&g ...

  6. ZJU 2676 Network Wars

    Network Wars Time Limit: 5000ms Memory Limit: 32768KB This problem will be judged on ZJU. Original I ...

  7. BZOJ 3774: 最优选择( 最小割 )

    最小割...二分染色然后把颜色不同的点的源汇反过来..然后就可以做了. 某个点(x,y): S->Id(x,y)(回报), Id(x,y)->T(代价), Id(i,j)&& ...

  8. 【BZOJ3774】最优选择 最小割

    [BZOJ3774]最优选择 Description 小N手上有一个N*M的方格图,控制某一个点要付出Aij的代价,然后某个点如果被控制了,或者他周围的所有点(上下左右)都被控制了,那么他就算是被选择 ...

  9. BZOJ 3774 最优选择 (最小割+二分图)

    题面传送门 题目大意:给你一个网格图,每个格子都有$a_{ij}$的代价和$b_{ij}$的回报,对于格子$ij$,想获得$b_{ij}$的回报,要么付出$a_{ij}$的代价,要么$ij$周围四联通 ...

随机推荐

  1. 2018.5.21 . XMLSpy激活的方法

    127.0.0.1 altova.com #XMLspy 127.0.0.1 www.altova.com #XMLspy 127.0.0.1 link.altova.com #XMLspy 追加加到 ...

  2. 高阶函数 -------JavaScript

    高阶函数 本文摘要:http://www.liaoxuefeng.com/ JavaScript的函数其实都指向某个变量.既然变量可以指向函数,函数的参数能接收变量,那么一个函数就可以接收另一个函数作 ...

  3. 基于Qt Creator实现中国象棋人机对战, c++实现

    GitHub地址: https://github.com/daleyzou/wobuku 这是自己大一学完c++后,在课程实践中写过的一个程序,实现象棋人机对战的算法还是有点难的, 自己当时差不多也是 ...

  4. numpy学习(二)

    ndarray的聚合操作 此博客讲的非常清楚,参照此博客即可 https://blog.csdn.net/qq_42571805/article/details/81146133

  5. 第2章 CentOS7集群环境配置

    目录 2.1 关闭防火墙 2.2 设置固定IP 2.3 修改主机名 2.4 添加用户 2.5 修改用户权限 2.6 新建目录 2.7 安装JDK 1.卸载系统自带的JDK 2.安装JDK 2.8 克隆 ...

  6. 初学js之qq聊天展开实例

    实现这样的效果. 直接看代码,html部分: <body> <div class="box"> <div class="lists" ...

  7. Python中函数参数类型和参数绑定

    参数类型 Python函数的参数类型一共有五种,分别是: POSITIONAL_OR_KEYWORD(位置参数或关键字参数) VAR_POSITIONAL(可变参数) KEYWORD_ONLY(关键字 ...

  8. SVM python小样例

    SVM有很多种实现,但是本章只关注其中最流行的一种实现,即序列最小化(SMO)算法在此之后,我们将介绍如何使用一种称为核函数的方式将SVM扩展到更多的数据集上基于最大间隔的分割数据优点:泛化错误率低, ...

  9. Apache Common-IO 使用

    Apache Common-IO 是什么? Apache File 工具类,能够方便的操作 File 运行环境 jdk 1.7 commons-io 2.6 测试代码 package com.m.ba ...

  10. C#开发微信公众平台教程

    http://www.cnblogs.com/xishuai/p/3625859.html http://www.cnblogs.com/wuhuacong/p/3614175.html http:/ ...