#1389 : Sewage Treatment

时间限制:2000ms
单点时限:2000ms
内存限制:256MB

描述

After years of suffering, people could not tolerate the pollution in the city any more, and started a long-lasting protest. Eventually, the municipal government made up its mind to deal with the sewage discharged by factories. The government planned to deploy some Pollution-Killer Utilities (PKUs) in order to clean all the sewage, and meanwhile, minimize the total cost.

The city can be approximately considered as a plane. There are n factories discharging sewage. The i-th one is located at (xi, yi), and discharges si units of sewage every day. The government has built m PKUs. The PKUs' locations are also fixed, and the j-th one is located at (xj, yj). Two features of PKUs are essential, and you will need to find the best choice of them: u is the upper limit of units of sewage that can be cleaned by one PKU every day, and c is the coverage radius of one PKU (which means, only if the distance between the PKU and the factory does not exceed the coverage radius, sewage discharged by this factory can be cleaned by this PKU). Note that all the PKUs share the same u and c. Because of some technical reasons, u has to be a positive integer.

Here is your task. The cost of deploying these PKUs can be weighed by an empirical formula:

f = u×sqrt(c)

You need to calculate the minimum value of f, and guarantee all the sewage is treated.

输入

There are no more than 15 test cases. Each test case starts with a line containing two positive integers n and m (1 <= n, m <= 100), representing the number of factories and PKUs. Then n lines follow, the i-th line contains three integers xi, yi, si (-10000 <= xi, yi <= 10000, 1 <= si <= 100), representing the i-th factory's location and the quantity of sewage discharged by it every day. In the next following m lines, the j-th line contains two integers xj, yj (-10000 <= xj, yj <= 10000), representing the j-th PKU's location. After each test case there is an empty line. After all the test cases there is a line "0 0", which indicates the end of the input, and should not be processed.

输出

For each test case, output a single line containing a number, which is the minimum value of f. This number should be rounded to an integer.

提示

Test case 1:

When u = 12 and c = 2, f has the minimum value of 12 * sqrt(2) = 16.97 = 17.

Test case 2:

When u = 10, c = sqrt(2), f has the minimum value of 10 * sqrt(sqrt(2)) = 11.89 = 12.

The input guarantees that there is always a valid solution. You may assume that the factories and PKUs are uniformly randomly distributed on the plane.

样例输入
3 1
-1 0 5
2 0 3
0 1 4
0 0 4 2
0 0 4
3 0 5
3 2 3
0 2 6
1 1
2 1 0 0
样例输出
17
12

题目链接:

  http://hihocoder.com/problemset/problem/1389

题目大意:

  N个污水厂fac,M个污水处理厂pku(N,M<=100)。每个污水厂坐标(xi,yi),会产生Si的污水。每个污水处理厂坐标(xi,yi),可以处理距离在C范围内的总量为U的污水(一个污水厂能被多个处理厂一起处理)。

  所有污水处理厂的U,C必须保持相同,求在能处理所有污水的情况下 f=U*√C 函数最小值。(U为正整数,C为欧拉距离,坐标范围在[-10000,10000])

题目思路:

  【二分+网络流+优化】

  首先求出每个污水处理厂b[i].from到每个污水厂b[i].to的欧拉距离b[i].dis,从小到大排序,总共N*M条边。设源S 汇T

  枚举C=b[i].dis(确保每个污水厂都和至少一个污水处理厂相连),二分需要的最小的U

  check时将前i条边都加入到网络流的图中,边的容量为无穷大。S到每个污水处理厂连一条容量为U的边,每个污水厂到T连一条容量为Si的边

    然后从S到T跑最大流看是否等于总污水量sum。等于就表明当前U可行,否则当前U不可行。

  光这样做会T,需要优化,二分U的上界为当前可行的最小的U(因为C在不断增大,f只有在U减小的情况下才可能更优),下界为sum/n,表示每个污水处理厂至少处理的均摊量。

  再加一个优化 当前枚举的C和二分下界l算出的函数f已经超过f的最优值时就break(再往后答案不会更优)。

  【我也不知道为什么这样就过了。比赛的时候觉得会T就没写。。结果看了大牛的题解发现真的是这么做。不过比赛的时候优化没想清楚可能还会T】

 //
//by coolxxx
//#include<bits/stdc++.h>
#include<iostream>
#include<algorithm>
#include<string>
#include<iomanip>
#include<map>
#include<stack>
#include<queue>
#include<set>
#include<bitset>
#include<memory.h>
#include<time.h>
#include<stdio.h>
#include<stdlib.h>
#include<string.h>
//#include<stdbool.h>
#include<math.h>
#define min(a,b) ((a)<(b)?(a):(b))
#define max(a,b) ((a)>(b)?(a):(b))
#define abs(a) ((a)>0?(a):(-(a)))
#define lowbit(a) (a&(-a))
#define sqr(a) ((a)*(a))
#define swap(a,b) ((a)^=(b),(b)^=(a),(a)^=(b))
#define mem(a,b) memset(a,b,sizeof(a))
#define eps (1e-10)
#define J 10000
#define mod 1000000007
#define MAX 0x7f7f7f7f
#define PI 3.14159265358979323
#pragma comment(linker,"/STACK:1024000000,1024000000")
#define N 204
#define M 20004
using namespace std;
typedef long long LL;
double anss;
LL aans,sum;
int cas,cass;
int n,m,lll,ans;
int C,U,S,T,nn;
int d[N],vd[N],last[N];
int mapp[N][N];
bool mark[N];
struct Point
{
int x,y,qq;
}fac[N],pku[N];
struct Edge
{
int from,to,dis;
}b[M];
struct xxx
{
int from,to,next,q;
}a[M<<];
bool cmp1(Point aa,Point bb)
{
if(aa.x!=bb.x)return aa.x<bb.x;
return aa.y<bb.y;
}
bool cmp2(Edge aa,Edge bb)
{
return aa.dis<bb.dis;
}
void add(int x,int y,int z)
{
a[++lll].to=y;
a[lll].q=z;
a[lll].next=last[x];
last[x]=lll;
}
int sap(int u,int f)
{
int i,v,tt,asp=,mix=nn-;
if(u==T)return f;
for(i=last[u];i!=;i=a[i].next)
{
v=a[i].to;
if(a[i].q>)
{
if(d[u]==d[v]+)
{
tt=sap(v,min(f-asp,a[i].q));
asp+=tt;
a[i].q-=tt;
a[i^].q+=tt;
if(asp==f || d[S]==nn)
return asp;
}
mix=min(mix,d[v]);
}
}
if(asp!=)return asp;
if(!--vd[d[u]])d[S]=nn;
else vd[d[u]=mix+]++;
return asp;
}
bool check(int u)
{
int i,f;
mem(d,);mem(vd,);
nn=T;
vd[]=nn;
ans=;
while(d[S]<nn)
{
f=sap(S,MAX);
ans+=f;
}
if(ans==sum)return ;
return ;
}
void build(int k,int uu)
{
int i;
mem(last,);lll=;
for(i=;i<=m;i++)
add(S,i,uu),add(i,S,);
for(i=;i<=n;i++)
add(m+i,T,fac[i].qq),add(T,m+i,);
for(i=;i<=k;i++)
add(b[i].from,b[i].to+m,MAX),add(b[i].to+m,b[i].from,);
}
int main()
{
#ifndef ONLINE_JUDGEW
// freopen("1.txt","r",stdin);
// freopen("2.txt","w",stdout);
#endif
int i,j,k;
int x,y,z,l,r,mid;
// init();
// for(scanf("%d",&cass);cass;cass--)
// for(scanf("%d",&cas),cass=1;cass<=cas;cass++)
// while(~scanf("%s",s))
while(~scanf("%d%d",&n,&m))
{
sum=;cas=;anss=1e30;ans=;
mem(mark,);
if(!n && !m)break;
for(i=;i<=n;i++)
{
scanf("%d%d%d",&fac[i].x,&fac[i].y,&fac[i].qq);
sum+=fac[i].qq;
}
for(i=;i<=m;i++)
scanf("%d%d",&pku[i].x,&pku[i].y);
//sort(fac+1,fac+1+n,cmp1);
//sort(pku+1,pku+1+m,cmp1);
for(i=;i<=m;i++)
{
for(j=;j<=n;j++)
{
b[++cas].from=i;
b[cas].to=j;
b[cas].dis=sqr(fac[j].x-pku[i].x)+sqr(fac[j].y-pku[i].y);
}
}
sort(b+,b++cas,cmp2);
S=n+m+,T=n+m+;
for(j=,k=;k<=cas;k++)
{
if(!mark[b[k].to])
{
j++,mark[b[k].to]=;
if(j==n)break;
}
}
U=sum;
for(;k<=cas;k++)
{
C=b[k].dis;
l=sum/n,r=U;
if((double)l*sqrt(sqrt(C))>anss)break;
while(l<r)
{
mid=(l+r)>>;
build(k,mid);
if(check(mid))r=mid;
else l=mid+;
}
U=min(U,r);
anss=min(anss,U*sqrt(sqrt(C)));
}
printf("%d\n",int(anss+0.5));
}
return ;
}
/*
// //
*/

hihoCoder 1389 Sewage Treatment 【二分+网络流+优化】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2016)网络赛)的更多相关文章

  1. hihoCoder 1391 Countries 【预处理+排序+堆】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2016)网络赛)

    #1391 : Countries 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 There are two antagonistic countries, countr ...

  2. hihoCoder 1392 War Chess 【模拟】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2016)网络赛)

    #1392 : War Chess 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 Rainbow loves to play kinds of War Chess gam ...

  3. hihoCoder 1582 Territorial Dispute 【凸包】(ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛)

    #1582 : Territorial Dispute 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 In 2333, the C++ Empire and the Ja ...

  4. hihoCoder 1584 Bounce 【数学规律】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛)

    #1584 : Bounce 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 For Argo, it is very interesting watching a cir ...

  5. hihoCoder 1578 Visiting Peking University 【贪心】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛)

    #1578 : Visiting Peking University 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 Ming is going to travel for ...

  6. hihoCoder 1586 Minimum 【线段树】 (ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛)

    #1586 : Minimum 时间限制:1000ms 单点时限:1000ms 内存限制:256MB 描述 You are given a list of integers a0, a1, …, a2 ...

  7. hihocoder 1586 ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛-题目9 : Minimum【线段树】

    https://hihocoder.com/problemset/problem/1586 线段树操作,原来题并不难..... 当时忽略了一个重要问题,就是ax*ay要最小时,x.y可以相等,那就简单 ...

  8. 【分类讨论】【计算几何】【凸包】hihocoder 1582 ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛 E. Territorial Dispute

    题意:平面上n个点,问你是否存在一种黑白染色方案,使得对于该方案,无法使用一条直线使得黑色点划分在直线一侧,白色点划分在另一侧.如果存在,输出一种方案. 如果n<=2,显然不存在. 如果所有点共 ...

  9. 【线段树】hihocoder 1586 ACM-ICPC国际大学生程序设计竞赛北京赛区(2017)网络赛 I. Minimum

    题意:给你一个序列(长度不超过2^17),支持两种操作:单点修改:询问区间中最小的ai*aj是多少(i可以等于j). 只需要线段树维护区间最小值和最大值,如果最小值大于等于0,那答案就是minv*mi ...

随机推荐

  1. 各大浏览器CSS Hack收集

    各大浏览器CSS Hack收集  >>>>>>>>>>>>>>>>>>>>> ...

  2. html获取gps坐标

    <script> function getLocation(){ var options={ enableHighAccuracy:true, maximumAge:1000 } if(n ...

  3. win7、xp下Meclipse SVN用户名修改

    Meclipse SVN用户名修改,在网上查找后发现如下方法: 1.查看你的Eclipse中使用的是什么SVNInterface windows>preference>Team>SV ...

  4. sql语句中like的使用

    先看一道题: 写出一条sql语句,找出表B中 字段Value中不全是字母 数字 下划线的数据 初看这道题,我们想到可以用like去进行模糊匹配,找出想要的结果.但是有一个地方需要注意:如果想在SQL ...

  5. Typescript 团队合作的利器

    前言 在介绍Typescript 之前,我需要隆重介绍一个人: 安德斯·海尔斯伯格(Anders Hejlsberg,1960.12~),丹麦人,Turbo Pascal编译器的主要作者,Delphi ...

  6. 【转】 NSString什么时候用copy,什么时候用strong

    原文: http://blog.csdn.net/itianyi/article/details/9018567 大部分的时候NSString的属性都是copy,那copy与strong的情况下到底有 ...

  7. jQuery 效果 - 淡入淡出

    通过 jQuery,您可以实现元素的淡入淡出效果. 点击展示 淡入/淡出 面板 实例 jQuery fadeIn()演示 jQuery fadeIn() 方法. jQuery fadeOut()演示 ...

  8. style currentStyle getComputedStyle的区别和用法

    先介绍下层叠样式表的三种形式: 1.内联样式,在html标签中style属性设置. <p style="color:#f90">内联样式</p> 2.嵌入样 ...

  9. java问题整理

    1.一个“.java”源文件中是否可以包括多个类(不是内部类)?有什么限制? 答:可以有多个类.但只能有一个public类.并且public类名必须与文件名相一致.   2.Java有没有goto? ...

  10. Tomcat设置最佳线程数总结

    最佳线程数: 性能压测的情况下,起初随着用户数的增加,QPS会上升,当到了一定的阀值之后,用户数量增加QPS并不会增加,或者增加不明显,同时请求的响应时间却大幅增加.这个阀值我们认为是最佳线程数. 为 ...