• Bzoj1901 Dynamic Ranking


    动态区间第k小

    离散化后

    那么每个点开一棵线段树(主席树)再套一个树状数组在外面

    每次询问区间内的树的个数时
    相当于进行了一次树状数组求区间和的操作,只是是把树状数组那个点看做主席树,对log棵主席树求区间和

    然后每次询问,修改时就是把log棵主席树同时跳到儿子,修改也是log棵

    时间复杂度O(nlogn*logn)空间复杂度O(nlogn*logn)

    # include <bits/stdc++.h>
    # define IL inline
    # define RG register
    # define Fill(a, b) memset(a, b, sizeof(a))
    using namespace std;
    typedef long long ll;
    const int _(2e4 + 10), __(2e6 + 10);
    
    IL ll Read(){
        RG char c = getchar(); RG ll x = 0, z = 1;
        for(; c < '0' || c > '9'; c = getchar()) z = c == '-' ? -1 : 1;
        for(; c >= '0' && c <= '9'; c = getchar()) x = (x << 1) + (x << 3) + (c ^ 48);
        return x * z;
    }
    
    int n, m, len, p[_], a[_], ql[_], qr[_], qk[_], qo[_];
    int ls[__], rs[__], sz[__], rt[__], num, tmp[2][20];
    
    IL void Modify(RG int &x, RG int l, RG int r, RG int pos, RG int val){
        if(!x) x = ++num;
        sz[x] += val;
        if(l == r) return;
        RG int mid = (l + r) >> 1;
        if(pos <= mid) Modify(ls[x], l, mid, pos, val);
        else Modify(rs[x], mid + 1, r, pos, val);
    }
    
    IL void PreModify(RG int x, RG int val){
        RG int k = lower_bound(p + 1, p + len + 1, a[x]) - p;
        for(RG int i = x; i <= n; i += i & -i) Modify(rt[i], 1, len, k, val);
    }
    
    IL int Query(RG int l, RG int r, RG int k){
        if(l == r) return l;
        RG int mid = (l + r) >> 1, sum = 0;
        for(RG int i = 1; i <= tmp[1][0]; i++) sum += sz[ls[tmp[1][i]]];
        for(RG int i = 1; i <= tmp[0][0]; i++) sum -= sz[ls[tmp[0][i]]];
        if(k <= sum){
            for(RG int i = 1; i <= tmp[1][0]; i++) tmp[1][i] = ls[tmp[1][i]];
            for(RG int i = 1; i <= tmp[0][0]; i++) tmp[0][i] = ls[tmp[0][i]];
            return Query(l, mid, k);
        }
        else{
            for(RG int i = 1; i <= tmp[1][0]; i++) tmp[1][i] = rs[tmp[1][i]];
            for(RG int i = 1; i <= tmp[0][0]; i++) tmp[0][i] = rs[tmp[0][i]];
            return Query(mid + 1, r, k - sum);
        }
    }
    
    IL int PreQuery(RG int l, RG int r, RG int k){
        Fill(tmp, 0);
        for(RG int i = r; i; i -= i & -i) tmp[1][++tmp[1][0]] = rt[i];
        for(RG int i = l - 1; i; i -= i & -i) tmp[0][++tmp[0][0]] = rt[i];
        return Query(1, len, k);
    }
    
    int main(RG int argc, RG char* argv[]){
        n = Read(); m = Read();
        for(RG int i = 1; i <= n; i++) a[i] = Read(), p[++len] = a[i];
        for(RG int i = 1; i <= m; i++){
            RG char c; scanf(" %c", &c);
            qo[i] = c == 'Q';
            if(qo[i]) ql[i] = Read(), qr[i] = Read(), qk[i] = Read();
            else ql[i] = qr[i] = Read(), qk[i] = Read(), p[++len] = qk[i];
        }
        sort(p + 1, p + len + 1); len = unique(p + 1, p + len + 1) - p - 1;
        for(RG int i = 1; i <= n; i++) PreModify(i, 1);
        for(RG int i = 1; i <= m; i++)
            if(qo[i]) printf("%d
    ", p[PreQuery(ql[i], qr[i], qk[i])]);
            else{
                PreModify(ql[i], -1);
                a[ql[i]] = qk[i];
                PreModify(ql[i], 1);
            }
        return 0;
    }
    
  • 相关阅读:
    损失函数及经验风险和结构风险
    ML面试题网站及ML模型网站
    数据归一化
    逻辑回归
    什么是凸函数及如何判断一个函数是否是凸函数
    Feature Selection Can Reduce Overfitting And RF Show Feature Importance
    Python Machine Learning-Chapter4
    skywalking与pinpoint全链路追踪方案对比
    全)Java从单体到微服务打造房产销售平台 2018年慕课网 高清视频+源码
    CentOS7 通过YUM安装MySQL5.7
  • 原文地址:https://www.cnblogs.com/cjoieryl/p/8206370.html
Copyright © 2020-2023  润新知