题目链接:

http://acm.hdu.edu.cn/showproblem.php?pid=4126

Genghis Khan the Conqueror

Time Limit: 10000/5000 MS (Java/Others)
Memory Limit: 327680/327680 K (Java/Others)
#### 问题描述
> Genghis Khan(成吉思汗)(1162-1227), also known by his birth name Temujin(铁木真) and temple name Taizu(元太祖), was the founder of the Mongol Empire and the greatest conqueror in Chinese history. After uniting many of the nomadic tribes on the Mongolian steppe, Genghis Khan founded a strong cavalry equipped by irony discipline, sabers and powder, and he became to the most fearsome conqueror in the history. He stretched the empire that resulted in the conquest of most of Eurasia. The following figure (origin: Wikipedia) shows the territory of Mongol Empire at that time.
> Our story is about Jebei Noyan(哲别), who was one of the most famous generals in Genghis Khan’s cavalry. Once his led the advance troop to invade a country named Pushtuar. The knights rolled up all the cities in Pushtuar rapidly. As Jebei Noyan’s advance troop did not have enough soldiers, the conquest was temporary and vulnerable and he was waiting for the Genghis Khan’s reinforce. At the meantime, Jebei Noyan needed to set up many guarders on the road of the country in order to guarantee that his troop in each city can send and receive messages safely and promptly through those roads.
>
> There were N cities in Pushtuar and there were bidirectional roads connecting cities. If Jebei set up guarders on a road, it was totally safe to deliver messages between the two cities connected by the road. However setting up guarders on different road took different cost based on the distance, road condition and the residual armed power nearby. Jebei had known the cost of setting up guarders on each road. He wanted to guarantee that each two cities can safely deliver messages either directly or indirectly and the total cost was minimal.
>
> Things will always get a little bit harder. As a sophisticated general, Jebei predicted that there would be one uprising happening in the country sooner or later which might increase the cost (setting up guarders) on exactly ONE road. Nevertheless he did not know which road would be affected, but only got the information of some suspicious road cost changes. We assumed that the probability of each suspicious case was the same. Since that after the uprising happened, the plan of guarder setting should be rearranged to achieve the minimal cost, Jebei Noyan wanted to know the new expected minimal total cost immediately based on current information.
#### 输入
> There are no more than 20 test cases in the input.
> For each test case, the first line contains two integers N and M (1
> The next line contains an integer Q (1 For each test case, output a real number demonstrating the expected minimal total cost. The result should be rounded to 4 digits after decimal point.
####样例输入
> 3 3
> 0 1 3
> 0 2 2
> 1 2 5
> 3
> 0 2 3
> 1 2 6
> 0 1 6
> 0 0

样例输出

6.0000

题意

给你一个无相图,要你求最小生成树,现在有q个查询,每个查询会改变一条边的权值,然后再改回去,要你求出每种情况下的最小生成树,最后求一下平均。

题解

1、如果改变的边不在我们一开始求的mst上,那么答案就是mst。

2、否则,把指定的树边删了,我们得到两个顶点集,那么代替原先那条边的一定是这两个集合的最短距离。所以我们只要处理出best[u][v](既u所在的集合和v所在的集合的最短距离)就能处理第2种情况了。做法是两次树dp,具体看代码。

dp[u][v]:顶点u到集合v的最短距离。

best[u][v]:集合u到集合v的最短距离。

代码

#include<map>
#include<set>
#include<cmath>
#include<queue>
#include<stack>
#include<ctime>
#include<vector>
#include<cstdio>
#include<string>
#include<bitset>
#include<cstdlib>
#include<cstring>
#include<iostream>
#include<algorithm>
#include<functional>
using namespace std;
#define X first
#define Y second
#define mkp make_pair
#define lson (o<<1)
#define rson ((o<<1)|1)
#define mid (l+(r-l)/2)
#define sz() size()
#define pb(v) push_back(v)
#define all(o) (o).begin(),(o).end()
#define clr(a,v) memset(a,v,sizeof(a))
#define bug(a) cout<<#a<<" = "<<a<<endl
#define rep(i,a,b) for(int i=a;i<(b);i++)
#define scf scanf
#define prf printf typedef long long LL;
typedef vector<int> VI;
typedef pair<int,int> PII;
typedef vector<pair<int,int> > VPII; const int INF=0x3f3f3f3f;
const LL INFL=0x3f3f3f3f3f3f3f3fLL;
const double eps=1e-8;
const double PI = acos(-1.0); //start---------------------------------------------------------------------- const int maxn=3030; struct Edge {
int u,v,w;
Edge(int u,int v,int w):u(u),v(v),w(w) {}
Edge() {}
bool operator < (const Edge& tmp) const {
return w<tmp.w;
}
} egs[maxn*maxn]; int n,m;
int G[maxn][maxn],dp[maxn][maxn],best[maxn][maxn];
bool used[maxn][maxn];
VPII tre[maxn]; int fa[maxn];
int find(int x) {
return fa[x]=fa[x]==x?x:find(fa[x]);
} void dfs(int u,int fa,int rt) {
if(u!=rt&&!used[u][rt]) dp[rt][u]=min(dp[rt][u],G[rt][u]);
rep(i,0,tre[u].sz()) {
int v=tre[u][i].X;
if(v==fa) continue;
dfs(v,u,rt);
dp[rt][u]=min(dp[rt][u],dp[rt][v]);
}
} int dfs2(int u,int fa,int rt) {
if(best[u][rt]<100000000) return best[u][rt];
best[u][rt]=min(best[u][rt],dp[u][rt]);
for(int i=0; i<tre[u].sz(); i++) {
int v=tre[u][i].X;
if(v==fa) continue;
dfs2(v,u,rt);
best[u][rt]=min(best[u][rt],best[v][rt]);
}
return best[u][rt];
} ///最小生成树
double kruskal() {
sort(egs,egs+m);
double mst=0;
for(int i=0; i<m; i++) {
int u=egs[i].u,v=egs[i].v,w=egs[i].w;
int pu=find(u);
int pv=find(v);
if(pu!=pv) {
fa[pv]=pu;
tre[u].pb(mkp(v,w));
tre[v].pb(mkp(u,w));
used[u][v]=used[v][u]=1;
mst+=w;
}
}
return mst;
} void init() {
clr(G,0x3f);
clr(dp,0x3f);
clr(best,0x3f);
clr(used,0);
for(int i=0; i<n; i++) fa[i]=i,tre[i].clear();
} int main() {
while(scf("%d%d",&n,&m)==2&&n) {
init();
for(int i=0; i<m; i++) {
int u,v,w;
scf("%d%d%d",&u,&v,&w);
G[u][v]=G[v][u]=w;
egs[i]=Edge(u,v,w);
} double mst=kruskal(); ///树形dp,dp[u][v]表示u到以v为根的子树的最短距离,既点到集合的距离(以u为根开始遍历,且只用非树边更新)
for(int i=0; i<n; i++) dfs(i,-1,i); ///记忆化搜索,best[u][v]表示以u为根的子树和以v为根的子树之间的最短距离,既集合到集合的距离
///对于子树v,我们已经求出了所有的点u到它的最短距离,现在只要遍历所有的u,求出最小值即可。
for(int i=0; i<n; i++) {
for(int j=0; j<tre[i].sz(); j++) {
int v=tre[i][j].X;
best[i][v]=best[v][i]=dfs2(i,v,v);
}
} int q;
scf("%d",&q);
double ans=0;
rep(i,0,q) {
int u,v,w;
scf("%d%d%d",&u,&v,&w);
if(used[u][v]) {
///在树边上
ans+=mst-G[u][v]+min(w,best[u][v]);
} else {
///不在树边上
ans+=mst;
}
} prf("%.4lf\n",ans/q);
}
return 0;
} //end-----------------------------------------------------------------------

HDU 4126 Genghis Khan the Conqueror 最小生成树+树形dp的更多相关文章

  1. HDU 4126 Genghis Khan the Conqueror MST+树形dp

    题意: 给定n个点m条边的无向图. 以下m行给出边和边权 以下Q个询问. Q行每行给出一条边(一定是m条边中的一条) 表示改动边权. (数据保证改动后的边权比原先的边权大) 问:改动后的最小生成树的权 ...

  2. hdu4126Genghis Khan the Conqueror (最小生成树+树形dp)

    Time Limit: 10000/5000 MS (Java/Others)    Memory Limit: 327680/327680 K (Java/Others) Total Submiss ...

  3. HDU 4126 Genghis Khan the Conqueror (树形DP+MST)

    题意:给一图,n个点,m条边,每条边有个花费,给出q条可疑的边,每条边有新的花费,每条可疑的边出现的概率相同,求不能经过原来可疑边 (可以经过可疑边新的花费构建的边),注意每次只出现一条可疑的边,n个 ...

  4. UVA- 1504 - Genghis Khan the Conqueror(最小生成树-好题)

    题意: n个点,m个边,然后给出m条边的顶点和权值,其次是q次替换,每次替换一条边,给出每次替换的边的顶点和权值,然后求出这次替换的最小生成树的值; 最后要你输出:q次替换的平均值.其中n<30 ...

  5. hdu4126Genghis Khan the ConquerorGenghis Khan the Conqueror(MST+树形DP)

    题目请戳这里 题目大意:给n个点,m条边,每条边权值c,现在要使这n个点连通.现在已知某条边要发生突变,再给q个三元组,每个三元组(a,b,c),(a,b)表示图中可能发生突变的边,该边一定是图中的边 ...

  6. HDU4126Genghis Khan the Conqueror(最小生成树+并查集)

    Genghis Khan the Conqueror Time Limit: 10000/5000 MS (Java/Others)    Memory Limit: 327680/327680 K ...

  7. HDU-4126 Genghis Khan the Conqueror 树形DP+MST (好题)

    题意:给出一个n个点m条边的无向边,q次询问每次询问把一条边权值增大后问新的MST是多少,输出Sum(MST)/q. 解法:一开始想的是破圈法,后来想了想应该不行,破圈法应该只能用于加边的情况而不是修 ...

  8. 刷题总结——Genghis Khan the Conqueror (hdu4126)

    题目: Genghis Khan(成吉思汗)(1162-1227), also known by his birth name Temujin(铁木真) and temple name Taizu(元 ...

  9. Install Air Conditioning HDU - 4756(最小生成树+树形dp)

    Install Air Conditioning HDU - 4756 题意是要让n-1间宿舍和发电站相连 也就是连通嘛 最小生成树板子一套 但是还有个限制条件 就是其中有两个宿舍是不能连着的 要求所 ...

随机推荐

  1. .NET中Quartz任务调度器的简单应用实例

    1.首先从NuGet中安装Quartz,安装最新版本就OK 2.新建一个Job类实现Quart中的IJob接口用于执行业务逻辑,代码如下: class CheckUpdateJob : IJob { ...

  2. Linux服务-ftp

    目录 1. ftp简介 2. ftp架构 3. ftp数据连接模式 4. 用户认证 5. vsftpd 5.1 vsftpd安装 5.2 vsftpd配置 5.3 vsftpd实验配置 5.3.1 安 ...

  3. Hadoop的HDFS和MapReduce的安装(三台伪分布式集群)

    一.创建虚拟机 1.从网上下载一个Centos6.X的镜像(http://vault.centos.org/) 2.安装一台虚拟机配置如下:cpu1个.内存1G.磁盘分配20G(看个人配置和需求,本人 ...

  4. 20155327Exp2 后门原理与实践

    20155327Exp2 后门原理与实践 一.实验说明 任务一:使用netcat获取主机操作Shell,cron启动 (0.5分) 任务二:使用socat获取主机操作Shell, 任务计划启动 (0. ...

  5. matplotlib雷达图

    用matplotlib画雷达图,网上流传的版本其实都是官网的一个例子.但是那个例子太复杂,而且它封装了几个类,让人难以一眼看出其本质. 我给出一个简单的解决方法,没有任何封装.作本文的原因,是为了回答 ...

  6. 【HNOI2015】亚瑟王

    题面 题解 考虑进行\(dp\). 设\(f[i][j]\)表示前\(i\)张卡中有\(j\)张被触发的概率. 我们可以知道第\(i\)张卡不被触发的概率为\((1 - p_i) ^ {r - j}\ ...

  7. COGS1871 [国家集训队2011]排队(魏铭)

    bzoj:http://www.lydsy.com/JudgeOnline/problem.php?id=2141 cogs:http://cogs.pro:8080/cogs/problem/pro ...

  8. .Net Core Linux centos7行—jenkins linux 构建.net core web app

    1.安装jdk.jenkins 是一个java web程序.所以必然需要jdk. yum install java 或者 yum install java-1.8.0-openjdk 2.下载jenk ...

  9. 并发系列(四)-----CAS

    一 简介 保证Java中的原子操做方式有两种方式  1 加锁(可以理解悲观锁机制)  2 CAS(可以理解为乐观锁机制)  CAS全称是Compare and Swap 即比较并替换.在JDK中许多地 ...

  10. Mac OS下Appium环境搭建及Genymotion模拟器安装

    说明:本机Mac本上已经安装的有:Python3.JDK.Node.js. Selenium Appium环境搭建 Step1:安装Appium Python Client库 1,命令:pip ins ...