斐波那契数列的O(logN)求法

欢迎来访

介绍求斐波那契数列时间复杂度为\(O(\log N)\)的做法之前,我们先看一下快速幂。

快速幂

题目链接

快速幂是数论中非常基础的算法。

当我们要求\(a^b mod p, (1 \le a, b, p \le 10^9)\)时,如果是朴素做法,时间复杂度为\(O(N)\)显然会超时,而快速幂能够做到的是将时间复杂度降到\(O(\log b)\)

做法

首先预处理出:\(a^{2^0}, a^{2^1}, a^{2^2}, a^{2^3}, ..., , a^{2^{\log b}}\)

将每一项相乘,可以得到:\(a^{2^0+2^1+2^2+2^3+...+2^{\log b}}\)

我们知道:\(2^0+2^1+2^2+2^3+...+2^{\log b}\)可以转换成二进制表示:\(1111...111\)一共有\(\log b + 1\)

利用\(2^i, 0 \le i \le \log b\)每一项选与不选可以凑出,\(0\) ~ \(2^{\log b + 1} - 1\)的任意整数。其中就包括我们要凑出的:\(b\)

这一步的时间复杂度为\(O(\log b)\)

C++代码

#include <iostream>
#include <cstdio>
#include <cstring>
#include <algorithm>

using namespace std;

typedef long long LL;

int qmi(int a, int b, int p) {
    LL res = 1 % p;
    while (b) {
        if (b & 1) res =  res * a % p;
        b >>= 1;
        a = (LL)a * a % p;
    }
    return res;
}

int main() {
    int n;
    int a, b, p;
    scanf("%d", &n);
    while (n--) {
        scanf("%d%d%d", &a, &b, &p);
        printf("%lld\n", qmi(a, b, p));
    }
    return 0;
}

斐波那契数列\(O(\log n)\)求法

首先先看一下斐波那契数列。

\[f(n) = \begin{cases} 1, & \text{$n = 1$} \\\\[2ex] 1, & \text{$n = 2$} \\\\[2ex] f(n-1)+f(n-2), & \text{$n \ge 2$} \end{cases} \]

我们设行向量\(F_n=[f_n, f_{n+1}]\),则:

\(F_1=[f_1, f_2]\)
\(F_2=[f_2, f_3]\)

我们看一下如何构造矩阵\(A\)使得\(F_1 \cdot A\)得到\(F_2\)

这个只要知道矩阵的乘法就不难构造出:

\(A=\begin{bmatrix} 0 & 1 \\ 1 & 1 \\ \end{bmatrix}\)

所以\(F_2=F1 \cdot A\)\(F_3=F2 \cdot A\),因为矩阵的乘法满足结合律,进而得到:

\(F_n=F_1\underbrace{A\cdot A\cdots A}_{\text{ n-1 times}}\),即\(F_n=F_1 \cdot A^{n-1}\)

这样我们就可以用快速幂来求了。

C++代码

#include <iostream>
#include <cstdio>
#include <cstring>
#include <algorithm>

using namespace std;

typedef long long LL;

const int MOD = 1e9 + 7;

LL res[2] = {1LL, 1LL};
LL A[2][2] = {
    {0LL, 1LL},
    {1LL, 1LL}
};

void mul(LL c[2], LL a[2], LL b[][2]) {
    
    LL tmp[2] = {0};
    for (int i = 0; i < 2; i++) 
        for (int j = 0; j < 2; j++)
            tmp[i] = tmp[i] + (a[j] * b[j][i]) % MOD;
    
    memcpy(c, tmp, sizeof tmp);
}

void mul(LL c[][2], LL a[][2], LL b[][2]) {
    
    LL tmp[2][2] = {0};
    for (int i = 0; i < 2; i++)
        for (int j = 0; j < 2; j++)
            for (int k = 0; k < 2; k++)
                tmp[i][j] = tmp[i][j] + (a[i][k] * b[k][j]) % MOD;
    
    memcpy(c, tmp, sizeof tmp);
}

LL fib(int n) {
    
    n--;
    while (n) {
        if (n & 1) mul(res, res, A);
        n >>= 1;
        mul(A, A, A);
    }
    
    return res[0];
}


int main() {
    
    int n;
    scanf("%d", &n);
    
    printf("%lld", fib(n));
    
    return 0;
}

拓展:求斐波那契前 n 项和\(O(\log n)\)

题目链接

分析

与上面的思路相同,在行向量中再加上和\(S_n\)

我们设行向量\(F_n=[f_n, f_{n+1}, S_n]\),则:

\(F_1=[f_1, f_2, S_1]\)
\(F_2=[f_2, f_3, S_2]\)

构造矩阵\(A\)使得\(F_1 \cdot A=F_2\),不难发现:

\(A=\begin{bmatrix} 0 & 1 & 0 \\\\ 1 & 1 & 1 \\\\ 0 & 0 & 1 \\\\ \end{bmatrix}\)

C++代码

#include <iostream>
#include <cstdio>
#include <cstring>
#include <algorithm>

using namespace std;

typedef long long LL;

int n, m;

int res[3] = {1, 1, 1};
int A[3][3] = {
    {0, 1, 0},
    {1, 1, 1},
    {0, 0, 1}
};

void mul(int c[3], int a[3], int b[][3]) {
    
    int tmp[3] = {0};
    for (int i = 0; i < 3; i++)
        for (int j = 0; j < 3; j++)
            tmp[i] = (tmp[i] + (LL)a[j] * b[j][i]) % m;
    
    memcpy(c, tmp, sizeof tmp);
}

void mul(int c[][3], int a[][3], int b[][3]) {
    
    int tmp[3][3] = {0};
    for (int i = 0; i < 3; i++)
        for (int j = 0; j < 3; j++)
            for (int k = 0; k < 3; k++)
                tmp[i][j] = (tmp[i][j] + (LL)a[i][k] * b[k][j]) % m;
                
    memcpy(c, tmp, sizeof tmp);
}
 
int main() {
    
    scanf("%d%d", &n, &m);
    
    n--;
    while (n) {
        if (n & 1) mul(res, res, A);
        mul(A, A, A);
        n >>= 1;
    }
    
    printf("%d", res[2]);
    
    return 0;
}

参考

AcWing蓝桥杯
求解斐波那契数列的若干方法

猜你喜欢

转载自www.cnblogs.com/optimjie/p/12707530.html