倍增法求LCA -- HihoCoder 1167 Advanced Theoretical Computer Science

#1167 : Advanced Theoretical Computer Science

Time Limit: 20000ms

Case Time Limit: 1000ms

Memory Limit: 256MB

Description

Yuuka is learning advanced theoretical computer science these days.But she cannot understand it. So she decided to hang out to relax.
Outside the door there is a tree of n nodes. She found there are P fairies on the tree. The fairies are very nervous, so each fairy only shows on a particular route on the tree. To be more specific, the i-th fairy will show on the path between nodes a_i and b_i.
Two fairies are friends if there route have some (no less than one) points in common.
Yuuka wants to know that how many pairs of fairies a,b (a!=b) are friends. Note that a,b and b,a are considered to be the same.

Input

The first line with two integers n and P (1 <= n, P <= 100000), meaning the size of the tree and the number of fairies. The nodes in the tree are numbered from 1 to n.
Then following n-1 lines with two integers a and b on each line, meaning there is an edge between a and b (a!=b).
Then following P lines, each line with two integers a_i and b_i, meaning the i-th fairy will show on the path from a_i to b_i. (a_i!=b_i)

Output

One line with an integer representing the answer.

Sample Input

6 3
1 2
2 3
2 4
4 5
4 6
1 3
1 5
5 6

Sample Output

2

题目的大意就是在一棵以1为根树上有若干个仙子。每个仙子都会在两个点之间游荡。如果两个仙子的游荡路径有重复，那么这两个仙子就是一对儿朋友。求总共有多少对朋友。

树 + 两点间路径，不难想到LCA。但是我们要如何记录朋友的数量呢?

经过观察不难发现：两个仙子互为朋友的充分必要条件是，其中一个仙子路径的LCA在另一个仙子的路径上。那么我们能否记录下每个节点作为LCA出现的次数，然后遍历每个仙子来积累朋友数量呢？

显然不行。因为很有可能两个仙子的LCA重合。在这样的情况下我们会把朋友数量算多，而且遍历的过程在数据量这么大的情况下铁定崩溃。既然蛋疼在重复计算上，或许我们可以分开讨论？

对于LCA重合的仙子们的朋友数量很容易求：设self[i]表示的是第i号节点作为LCA的次数。那么这种朋友的数目总和就是[sigma(1,n)](self[i] * (self[i] - 1]) / 2)。

（第一次写博客不知道那个求和的sigma咋打出来，难看了点，见谅！）

简单的情况考虑完了，现在我们要计算“LCA在路径上，且两个LCA不能重合”的朋友对数（我们称之为B类朋友）。如果定义sum[i]表示第i号节点，及其所有长辈节点的self[i]的数量和，那么对于一个穿梭于a和b之间的仙子来说，和它互为B类朋友的人数为sum[a] + sum[b] - 2 * sum[LCA[a][b]] 。 LCA那里被计算了两次，因此要全部减出去。

写到这里，估计有朋友会问：那又何必分成两类？直接把所有朋友都当作B类朋友，然后在计算的时候减去一倍的sum[LCA[a][b]]不就完了么？

其实我第一次就是这么搞的。幸亏没过样例- -如果这样的话，假设a和b两个仙子的LCA相同。在计算a的朋友的时候我们把b算进去了。在计算b的时候不是又把a也给算进去了吗？所以说，LCA相同的情况本身就是一个蛋疼的情况，需要特殊考虑一下。

#include<iostream>
#include<stdio.h>
#include<string.h>
#include<algorithm>
#include<list>
#include<vector>
#include<stdlib.h>
#define N 100005
#define LOGN 20
using namespace std;

vector<int>e[N];
int sum[N],self[N],parent[LOGN][N],depth[N],n,m,p;
long long int ans = 0;//要注意数据范围。因为没用long long我还哇了一发。
pair<int,int>f[N];//ÏÉ×Ó 

/***************LCA**************/

void dfs(int v,int p,int d)
{
	parent[0][v] = p;
	depth[v] = d;
	for(int i = 0;i<e[v].size();i++)
	if(e[v][i] != p)dfs(e[v][i],v,d+1);
}

int lca(int u,int v)
{
	if(depth[u] > depth[v])swap(u,v);
	for(int k = 0; k < LOGN;k++)
	if(depth[v] - depth[u] >> k & 1)
	v = parent[k][v];
	if(u == v)
	return u;
	for(int k = LOGN - 1;k >= 0;k--)
	if(parent[k][u] != parent[k][v])
	u = parent[k][u],v = parent[k][v];
	return parent[0][u];
}

/***************LCA**************/

void getSum(int v,int p)
{
	sum[v] = self[v] + sum[p];
	for(int i = 0;i<e[v].size();i++)
	if (e[v][i] != p) getSum(e[v][i],v);
}

void init()
{
	int root = 1;
	int t1,t2;
	for(int i = 1;i<=m;i++)
	{
		cin>>t1>>t2;
		e[t1].push_back(t2);
		e[t2].push_back(t1);
	}
	dfs(root,-1,0);
	for(int k = 0; k+ 1 < LOGN;k++)
	for(int v = 1; v <= n; v++)
	if(parent[k][v] < 0)parent[k+1][v] = -1;
	else parent[k+1][v] = parent[k][parent[k][v]];
	for(int i = 1;i<=p;i++)
	{
		cin>>f[i].first>>f[i].second;
		int h = lca(f[i].first,f[i].second);
		self[h]++;
	}
	getSum(root,0);
}

void work()
{
	for(int i = 1;i<=p;i++)
		ans += sum[f[i].first] + sum[f[i].second] - 2 * sum[lca(f[i].first,f[i].second)];
	for(int i = 1;i<=n;i++)
		ans += (long long int)self[i] * (long long int)(self[i] - 1) / 2;
}

int main()
{
	cin>>n>>p;
	m = n - 1;//懒得改模板了，表喷我
	init();
	work();
	cout<<ans<<endl;
	return 0;
}