标签:upd tree return blog opened sigma str struct 思路
离散化+树状数组+排列组合
很久以前就看到过这道题,现在依然不会做。。。看完题解发现思路很简单,就是有点难写
我们先将坐标离散化,x和y最大是w,然后我们就有了一个暴力做法, 枚举每块墓地,统计,因为墓地上下左右没东西的话就不可能有贡献,这些坐标自然就被离散化了,所以墓地最多有w*w块
复杂度O(w*w),然后我们优化一下,发现不用枚举每个墓地,对于相邻两棵树,他们中间墓地左右的树的数量是相同的,这样的方案有C(l,k)*C(r,k),对于每个墓地,这样的虔诚度是C(l,k)*C(r,k)*C(up,k)*C(down,k),我们发现两棵树中间的墓地C(l,k)*C(r,k)是相同的,现在需要的就是统计sigma(C(up,k)*C(down,k)),我们可以用树状数组维护这个东西,先把墓地按行排序分类,然后每次对于每行枚举相邻的两棵树,计算之间的sigma,这样用树状数组统计,然后就是更新,有树的地方是需要更新的,就是把对应的y坐标离散化后的位置在树状数组里更新,减去原来的值,改成现在新的C(up,k)*C(down,k)
#include<bits/stdc++.h> using namespace std; const int N = 100010; const long long mod = 2147483648ll; int n, w, m, k; int up[N], down[N]; long long c[N][12], BIT[N], ans; struct Tree { int x, y, ux, uy, left, right, up, down; bool friend operator < (Tree A, Tree B) { return A.ux == B.ux ? A.uy < B.uy : A.ux < B.ux; } } tree[N]; vector<Tree> line[N]; inline int lowbit(int i) { return i & (-i); } inline void update(int pos, int delta) { for(int i = pos; i <= w + 5; i += lowbit(i)) BIT[i] = (BIT[i] + delta) % mod; } inline long long query(int pos) { long long ret = 0; for(int i = pos; i; i -= lowbit(i)) ret = (ret + BIT[i]) % mod; return ret; } int main() { scanf("%d%d%d", &n, &m, &w); vector<int> vx, vy; for(int i = 1; i <= w; ++i) { scanf("%d%d", &tree[i].x, &tree[i].y); vx.push_back(tree[i].x); vy.push_back(tree[i].y); } scanf("%d", &k); c[0][0] = 1; for(int i = 1; i <= w; ++i) { c[i][0] = 1; for(int j = 1; j <= min(k, i); ++j) c[i][j] = (c[i - 1][j] + c[i - 1][j - 1]) % mod; } sort(vx.begin(), vx.end()); sort(vy.begin(), vy.end()); vx.erase(unique(vx.begin(), vx.end()), vx.end()); vy.erase(unique(vy.begin(), vy.end()), vy.end()); for(int i = 1; i <= w; ++i) { tree[i].ux = lower_bound(vx.begin(), vx.end(), tree[i].x) - vx.begin() + 1; tree[i].uy = lower_bound(vy.begin(), vy.end(), tree[i].y) - vy.begin() + 1; // printf("ux=%d uy=%d\n", tree[i].ux, tree[i].uy); } sort(tree + 1, tree + w + 1); for(int i = 1; i <= w; ++i) { ++up[tree[i].uy]; tree[i].up = up[tree[i].uy]; } for(int i = w; i; --i) { tree[i].down = down[tree[i].uy]; ++down[tree[i].uy]; } for(int i = 1; i <= w; ++i) line[tree[i].ux].push_back(tree[i]); for(int i = 1; i <= vx.size(); ++i) { sort(line[i].begin(), line[i].end()); for(int j = 0; j < line[i].size(); ++j) { Tree &tmp = line[i][j]; tmp.left = j + 1; tmp.right = line[i].size() - j; } } for(int i = 1; i <= vx.size(); ++i) { for(int j = 1; j < line[i].size(); ++j) { Tree tmp_l = line[i][j - 1], tmp_r = line[i][j]; long long comb_left = c[tmp_l.left][k], comb_right = c[tmp_r.right][k], sigma = query(tmp_r.uy - 1) - query(tmp_l.uy); ans = (ans + comb_left % mod * comb_right % mod * sigma % mod) % mod; } for(int j = 0; j < line[i].size(); ++j) { Tree tmp = line[i][j]; update(tmp.uy, c[tmp.up][k] % mod * c[tmp.down][k] % mod - query(tmp.uy) + query(tmp.uy - 1)); } } printf("%lld\n", (ans % mod + mod) % mod); return 0; }
标签:upd tree return blog opened sigma str struct 思路
原文地址:http://www.cnblogs.com/19992147orz/p/7337355.html