Codeforces 536F Lunar New Year and a Recursive Sequence | BSGS/exgcd/矩阵乘法

我诈尸啦!

高三退役选手好不容易抛弃天利和金考卷打场CF,结果打得和shi一样……还因为queue太长而unrated了!一个学期不敲代码实在是忘干净了……

没分该没分,考题还是要订正的 =v= 欢迎阅读本题解!

P.S. 这几个算法我是一个也想不起来了 TAT

题目链接

Codeforces 536F Lunar New Year and a Recursive Sequence 新年和递推数列

题意描述

某数列\(\{f_i\}\)递推公式:\[f_i = (\prod_{j=1}^kf_{i-j}^{b_j}) \bmod p\]

其中\(b\)是已知的长度为\(k\)的数列,\(p = 998244353\),\(f_1 = f_2 = ... = f_{k-1} = 1\),\(f_k\)未知。

给出两个数\(n, m\),构造一个\(f_k\)使得\(f_n = m\),无解输出-1。

\(k \le 100, n \le 10^9\)

题解

数论!真令人头秃!

首先这个数据范围让人想到什么?矩阵乘法!

矩阵乘法想推这个全是乘法和乘方的递推数列咋办?取对数!离散对数!

于是这道题关键的两个考点就被你发现啦!

(然而我太菜了,并不能发现 = =)

什么是离散对数?

众所周知(众==学过NTT的人等),这个喜闻乐见的模数\(p = 998244353\)有个原根\(g=3\),\(g^i(0\le i < P - 1)\)和\(1\le x < P\)一一对应。那么类比我们学过的对数,称这个\(i\)为\(x\)的离散对数。

令数列\(h_i\)为\(f_i\)的离散对数。

那么有递推式:\[h_i = (\sum_{j=1}^kb_j\cdot h_{i-j}) \bmod (p - 1)\]

其中\(h_1 = h_2 = ... = h_{k-1} = 0\)。注意模数变成了\(p - 1\)(费马小定理)。

这个就可以用矩阵加速了!如果我们把\(h_k\)设为1带进去,求得\(f_n = c\),那么有\(h_n = c \cdot h_k \bmod (p - 1)\);

\(h_n\)即为\(m\)的离散对数,用BSGS可求;

exgcd解刚才这个同余方程即可得到\(h_k\);

\(f_k = g^{h_k}\),快速幂即可得到\(f_k\)。

如果exgcd发现没有解的话就输出-1。

是不是思路非常清晰啊~

代码

#include <cstdio>
#include <cstring>
#include <cmath>
#include <algorithm>
#include <iostream>
#include <cassert>
#define space putchar(' ')
#define enter putchar('\n')
using namespace std;
typedef long long ll;
template <class T>
void read(T &x){
    char c;
    bool op = 0;
    while(c = getchar(), c < '0' || c > '9')
        if(c == '-') op = 1;
    x = c - '0';
    while(c = getchar(), c >= '0' && c <= '9')
        x = x * 10 + c - '0';
    if(op) x = -x;
}
template <class T>
void write(T x){
    if(x < 0) putchar('-'), x = -x;
    if(x >= 10) write(x / 10);
    putchar('0' + x % 10);
}

const int N = 102, P = 998244353, P2 = 998244352, G = 3;
int K;
ll b[N], n, m, C;

namespace BSGS {
    const int S = 32000, M = 2000000;
    int cnt = 0, adj[M + 5], nxt[S + 5];
    ll key[S + 5], val[S + 5];
    void insert(ll K, ll V){
        int p = K % M;
        key[++cnt] = K;
        val[cnt] = V;
        nxt[cnt] = adj[p];
        adj[p] = cnt;
    }
    ll search(ll K){
        for(int u = adj[K % M]; u; u = nxt[u])
            if(key[u] == K) return val[u];
        return -1;
    }
    void init(){
        ll sum = 1;
        for(int i = 1; i <= S; i++)
            sum = sum * G % P;
        ll tot = 1;
        for(int i = 1; (i - 1) * S < P - 1; i++)
            tot = tot * sum % P, insert(tot, i * S);
    }
    ll log(ll x){
        ll sum = 1, ret;
        for(int i = 1; i <= S; i++){
            sum = sum * G % P;
            ret = search(sum * x % P);
            if(~ret && ret < P - 1) return ret - i;
        }
        assert(0);
        return -1;
    }
}

struct matrix {
    ll g[N][N];
    matrix(){
        memset(g, 0, sizeof(g));
    }
    matrix(int x){
        memset(g, 0, sizeof(g));
        for(int i = 1; i <= K; i++)
            g[i][i] = 1;
    }
    matrix operator * (const matrix &b){
        matrix c;
        for(int i = 1; i <= K; i++)
            for(int j = 1; j <= K; j++)
                for(int k = 1; k <= K; k++)
                    c.g[i][j] = (c.g[i][j] + g[i][k] * b.g[k][j]) % P2;
        return c;
    }
};

ll qpow(ll a, ll x){
    ll ret = 1;
    while(x){
        if(x & 1) ret = ret * a % P;
        a = a * a % P;
        x >>= 1;
    }
    return ret;
}
matrix qpow(matrix a, ll x){
    matrix ret(1);
    while(x){
        if(x & 1) ret = ret * a;
        a = a * a;
        x >>= 1;
    }
    return ret;
}
ll calcC(){
    matrix ret, op;
    ret.g[K][1] = 1;
    for(int i = 1; i < K; i++)
        op.g[i][i + 1] = 1;
    for(int i = 1; i <= K; i++)
        op.g[K][i] = b[K - i + 1];
    ret = qpow(op, n - K) * ret;
    return ret.g[K][1];
}
void exgcd(ll a, ll b, ll &g, ll &x, ll &y){
    if(!b) return (void)(x = 1, y = 0, g = a);
    exgcd(b, a % b, g, y, x);
    y -= x * (a / b);
}
ll solve(ll A, ll B){ //Ax % P2 == B, solve x
    ll a = A, b = P2, g, x, y;
    exgcd(a, b, g, x, y);
    if(B % g) return -1;
    x *= B / g, y *= B / g;
    ll t = b / g;
    x = (x % t + t) % t;
    return x;
}

int main(){

    BSGS::init();
    read(K);
    for(int i = 1; i <= K; i++) read(b[i]);
    read(n), read(m);
    C = calcC();
    m = BSGS::log(m);
    ll ans = solve(C, m);
    if(ans == -1) puts("-1");
    else write(qpow(G, ans)), enter;

    return 0;
}

原文地址:https://www.cnblogs.com/RabbitHu/p/10349388.html

时间: 2024-10-07 21:32:45

Codeforces 536F Lunar New Year and a Recursive Sequence | BSGS/exgcd/矩阵乘法的相关文章

Codeforces 1106F Lunar New Year and a Recursive Sequence (数学、线性代数、线性递推、数论、BSGS、扩展欧几里得算法)

哎呀大水题..我写了一个多小时..好没救啊.. 数论板子X合一? 注意: 本文中变量名称区分大小写. 题意: 给一个\(n\)阶递推序列\(f_k=\prod^{n}_{i=1} f_{k-i}b_i\mod P\)其中\(P=998244353\), 输入\(b_1,b_2,...,b_n\)以及已知\(f_1,f_2,...,f_{n-1}=1\), 再给定一个数\(m\)和第\(m\)项的值\(f_m\), 求出一个合法的\(f_n\)值使得按照这个值递推出来的序列满足第\(m\)项的值为

CF1106F Lunar New Year and a Recursive Sequence(矩阵快速幂+bsgs+exgcd)

题面 传送门 前置芝士 \(BSGS\) 什么?你不会\(BSGS\)?百度啊 原根 对于素数\(p\)和自然数\(a\),如果满足\(a^x\equiv 1\pmod{p}\)的最小的\(x\)为\(p-1\),那么\(a\)就是\(p\)的一个原根 离散对数 对于素数\(p\),以及\(p\)的一个原根\(g\),定义\(y\)为\(x\)的离散对数,当且仅当\(g^y\equiv x\pmod{p}\),记\(y\)为\(ind_g x\).不难发现原数和离散对数可以一一对应.也不难发现离

CF1106F Lunar New Year and a Recursive Sequence

又傻掉了呢 看到连乘显然直接转原根变成线性齐次递推式. 矩阵乘法求一发. 然后分析一下发现是个x^k=m的形式. 按照套路解一下高次方程就好了. 需要用到exgcd和bsgs. #include<iostream> #include<cctype> #include<cstdio> #include<cstring> #include<string> #include<cmath> #include<ctime> #inc

CF1106F Lunar New Year and a Recursive Sequence——矩阵快速幂&amp;&amp;bsgs

题意 设 $$f_i = \left\{\begin{matrix}1 , \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \  i < k\\ \prod_{j=1}^k f_{i-j}^{b_j} \ mod \ p, \ \ \ \ \ i > k\end{matrix}\right.$$ 求 $f_k$($1 \leq f_k < p$),使得 $f_m = n$.($1 \leq k\leq 100$) 分析 $f_n$ 可以表示

Recursive sequence(HDU5950 矩阵快速幂)

题目: Farmer John likes to play mathematics games with his N cows. Recently, they are attracted by recursive sequences. In each turn, the cows would stand in a line, while John writes two positive numbers a and b on a blackboard. And then, the cows wou

HDu 5950 Recursive sequence(矩阵快速幂)

Recursive sequence Time Limit: 2000/1000 MS (Java/Others)    Memory Limit: 65536/65536 K (Java/Others)Total Submission(s): 1323    Accepted Submission(s): 589 Problem Description Farmer John likes to play mathematics games with his N cows. Recently,

HDU 5950 Recursive sequence 【递推+矩阵快速幂】 (2016ACM/ICPC亚洲区沈阳站)

Recursive sequence Time Limit: 2000/1000 MS (Java/Others)    Memory Limit: 65536/65536 K (Java/Others)Total Submission(s): 249    Accepted Submission(s): 140 Problem Description Farmer John likes to play mathematics games with his N cows. Recently, t

Recursive sequence HDU - 5950

Recursive sequence HDU - 5950 题意:求 f(n) = f(n?1)+2*f(n?2)+n4,其中 f(1)=a,f(2)=b 利用矩阵加速~ 比较坑的是mod=2147493647....不是2147483647,用int WA了多次=_=|| 1 #include <bits/stdc++.h> 2 using namespace std; 3 #define ll long long 4 const ll mod=2147493647; 5 const int

HDU 4914 Linear recursive sequence(矩阵乘法递推的优化)

题解见X姐的论文 矩阵乘法递推的优化,只是mark一下.. HDU 4914 Linear recursive sequence(矩阵乘法递推的优化)