算法学习————FWT

解决问题

\[C_i = \sum\limits_{j\bigoplus k = i} A_j\times B_k \]

其中\(\bigoplus\)为or,and,xor,已知A和B,求解C

和FFT还有NTT的思想都是一样的,考虑在FFT的时候,我们是从系数法转化成点值法

对A和B本身FFT一次,想乘后得到C,然后用逆运算再把点值法转化成系数法,下面FWT也是一样的流程

Or

直接设\(FWTA(i) = \sum\limits_{j|i = i} A_j\)

假设可以通过\(FWTA(i)\)和\(FWTB(i)\)想乘得到\(FWTC(i)\),有\(FWTA(i)\times FWTB(i) = FWTC(i)\)

如何证明这个式子?

左边:

\[FWTA(i)\times FWTB(i) = \sum\limits_{j|i = i} A_j\sum\limits_{k|i = i} B_k \]

\[= \sum\limits_{j|i = i}\sum\limits_{k|i = i} A_j\times B_k \]

\[= \sum\limits_{j|k|i = i}A_j\times B_k \]

这里解释一下为什么j|i = i且k|i = i可以合并为j|k|i = i,我们可以知道或取得是两个数1的并集

这就说明j的1是i的1的子集,同理k,那么j和k的子集也是i的子集

右边:

\[FWTC(i) = \sum\limits_{l|i = i} C_i \]

\[= \sum\limits_{l|i = i} \sum\limits_{x|y = l} A_i\times B_j \]

\[= \sum\limits_{x|y|i = i}A_x\times B_y \]

易证两边等价

And

对于与也是一样的,设\(FWTA(i) = \sum\limits_{j \And i = i} A_i\)

左边:

\[FWTA(i)\times FWTB(i) = \sum\limits_{j \And i = i}A_j\sum\limits_{k\And i = i} B_k \]

\[= \sum\limits_{j\And i = i}\sum\limits_{k\And i = i} A_j\times B_k \]

\[= \sum\limits_{j\And i\And i = i} A_j\times B_k \]

右边:

\[FWTC(i) = \sum\limits_{l\And i = i} C_i \]

\[= \sum\limits_{l\And i = i} \sum\limits_{x\And y = l} A_i\times B_j \]

\[= \sum\limits_{x\And y\And i = i}A_x\times B_y \]

Xor

比较麻烦的就是异或操作了

设\(FWTA(i) = \sum\limits_{j} A_j\times (-1)^{count(i\And j)}\)

那么左边:

\[FWTA(i)\times FWRB(i) = \sum\limits_{j} A_j\times (-1)^{count(i\And j)}\sum\limits_{k} B_k\times (-1)^{count(i\And k)} \]

\[= \sum\limits_j \sum\limits_k A_j\times (-1)^{count(i\And j)}\times b_k\times (-1)^{count(i\And k)} \]

\[= \sum\limits_j \sum\limits_k A_j\times B_k\times (-1)^{count(i\And j)+count(i\And k)} \]

右边:

\[FWTC(i) = \sum\limits_{l} C_l\times (-1)^{count(i\And l)} \]

\[= \sum\limits_l [(-1)^{count(i\And (x\bigoplus y))} \sum\limits_{x\bigoplus y = l} A_x\times B_y] \]

\[= \sum\limits_x \sum\limits_y A_x\times B_y\times (-1)^{count(i\And (x\bigoplus y))} \]

观察两边的形式,那么我们现在只需要证明\((-1)^{count(i\And (x\bigoplus y))} = (-1)^{count(i\And x)+count(i\And y)}\)

我们可以知道左边异或操作把x和y共有的1给消掉了,而右边对于x和y共有的1,\((-1)^2 = 1\)乘1没有贡献,剩下的都是x和y不共有的1产生的贡献,数量相同

式子到这里就推完了,那么现在的问题就是如果快速的求出\(FWTA(i) FWTB(i)\),以及求逆的过程

我们把序列分成几段,刚开始长度为2:

000 001 | 010 011 | 100 101 | 110 111

先看对于或操作,我们可以知道一个数,包含的1是他的1的子集的数会对他产生贡献,就看块内前一个数会对后一个数产生贡献,贡献累加

对于与操作,则反过来,后一个数对前一个数有贡献,对于异或操作,好像可以推式子得来

之后我们可以每次把块长$\times$2,贡献一直累加,就能求出最后的答案

代码:

#include <iostream>
#include <cstring>
#include <algorithm>
#include <cstdio>
#define int long long
#define O(x) cout<<#x<<" "<<x<<endl;
#define o(x) cout<<#x<<" "<<x<<" ";
using namespace std;
int read(){
	int x = 1,a = 0;char ch = getchar();
	while (ch < '0'||ch > '9'){if (ch == '-') x = -1;ch = getchar();}
	while (ch >= '0'&&ch <= '9'){a = a*10+ch-'0';ch = getchar();}
	return x*a;
}
const int maxn = 2e6+10,mod = 998244353;
int n,inv;
int A[maxn],B[maxn];
int qpow(int a,int k){
	int res = 1;a = a % mod;
	while (k){
		if (k&1) res = res*a % mod;
		a = a*a % mod;
		k >>= 1;
	}
	return res % mod;
}
int a[maxn],b[maxn];
void FWTOr(int a[],int op){
	for (int l = 1;l < n;l <<= 1){
		for (int i = 0;i < n;i += (l << 1)){
			for (int j = 0;j < l;j++){
				(a[i+j+l] += op*a[i+j]) %= mod;
			}
		}
	}
}
void FWTAnd(int a[],int op){
	for (int l = 1;l <= n;l <<= 1){
		for (int i = 0;i < n;i += (l << 1)){
			for (int j = 0;j < l;j++){
				(a[i+j] += op*a[i+j+l]) %= mod;
			}
		}
	}
}
void FWTXor(int a[],int op){
	for (int l = 1;l < n;l <<= 1){
		for (int i = 0;i < n;i += (l << 1)){
			for (int j = 0;j < l;j++){
				int x = a[i+j],y = a[i+j+l];
				a[i+j] = (x+y) % mod,a[i+j+l] = (x+mod-y) % mod;
				if (op == -1) a[i+j] = a[i+j]*inv % mod,a[i+j+l] = a[i+j+l]*inv % mod;
			}
		}
	}
}
signed main(){
	n = read(),n = (1 << n);
	inv = qpow(2,mod-2);
	for (int i = 0;i < n;i++) a[i] = read();
	for (int i = 0;i < n;i++) b[i] = read();
	for (int i = 0;i < n;i++) A[i] = a[i],B[i] = b[i];
	FWTOr(A,1),FWTOr(B,1);
	for (int i = 0;i < n;i++) A[i] = A[i]*B[i] % mod;
	FWTOr(A,-1);
	for (int i = 0;i < n;i++) cout<<((A[i] + mod) % mod)<<" ";
	cout<<endl;
	for (int i = 0;i < n;i++) A[i] = a[i],B[i] = b[i];
	FWTAnd(A,1),FWTAnd(B,1);
	for (int i = 0;i < n;i++) A[i] = A[i]*B[i] % mod;
	FWTAnd(A,-1);
	for (int i = 0;i < n;i++) cout<<((A[i] + mod) % mod)<<" ";
	cout<<endl;
	for (int i = 0;i < n;i++) A[i] = a[i],B[i] = b[i];
	FWTXor(A,1),FWTXor(B,1);
	for (int i = 0;i < n;i++) A[i] = A[i]*B[i] % mod;
	FWTXor(A,-1);
	for (int i = 0;i < n;i++) cout<<((A[i] + mod) % mod)<<" ";
	cout<<endl;
	return 0;
}
上一篇:【瞎口胡】快速傅里叶变换 / FFT


下一篇:【ybt金牌导航5-4-4】【luogu P4842】城市旅行