您的位置:首页 > 运维架构 > Shell

HDU5730 Shell Necklace(DP + CDQ分治 + FFT)

2016-09-19 18:40 260 查看

题目

Source

http://acm.hdu.edu.cn/showproblem.php?pid=5730

Description

Perhaps the sea‘s definition of a shell is the pearl. However, in my view, a shell necklace with n beautiful shells contains the most sincere feeling for my best lover Arrietty, but even that is not enough.

Suppose the shell necklace is a sequence of shells (not a chain end to end). Considering i continuous shells in the shell necklace, I know that there exist different schemes to decorate the i shells together with one declaration of love.

I want to decorate all the shells with some declarations of love and decorate each shell just one time. As a problem, I want to know the total number of schemes.

Input

There are multiple test cases(no more than 20 cases and no more than 1 in extreme case), ended by 0.

For each test cases, the first line contains an integer n, meaning the number of shells in this shell necklace, where 1≤n≤105. Following line is a sequence with n non-negative integer a1,a2,…,an, and ai≤107 meaning the number of schemes to decorate i continuous shells together with a declaration of love.

Output

For each test case, print one line containing the total number of schemes module 313(Three hundred and thirteen implies the march 13th, a special and purposeful day).

Sample Input

3
1 3 7
4
2 2 2 2
0

Sample Output

14
54

分析

题目大概说已知连续i(1<=i<=n)个贝壳组合成一段项链的方案数a[i],求组合成包含n个贝壳的项链的总方案数。

dp[i]表示组合成包含i个贝壳的项链的总方案数

转移:dp[i]=Σdp[i-j]*a[j](1<=j<=i)

直接枚举转移的话时间复杂度O(n2),是不行的。

其实这个转移方程是个比较特殊的卷积形式,可以用FFT去求,但是从1到n依次求的话时间复杂度是O(n2logn)。

而利用CQD分治,每次治的过程累加左半区间内各个已经求得dp值的状态对右半区间各个状态的贡献,这个贡献就是那个方程用FFT求即可,这样时间复杂度由主定理可知是O(nlog2n)。

这是个很经典的题目吧,虽然现在才做。。

代码

#include<cstdio>
#include<cstring>
#include<cmath>
#include<algorithm>
using namespace std;
#define INF (1<<30)
#define MAXN 333333
const double PI=acos(-1.0);

struct Complex{
double real,imag;
Complex(double _real,double _imag):real(_real),imag(_imag){}
Complex(){}
Complex operator+(const Complex &cp) const{
return Complex(real+cp.real,imag+cp.imag);
}
Complex operator-(const Complex &cp) const{
return Complex(real-cp.real,imag-cp.imag);
}
Complex operator*(const Complex &cp) const{
return Complex(real*cp.real-imag*cp.imag,real*cp.imag+cp.real*imag);
}
void setValue(double _real=0,double _imag=0){
real=_real; imag=_imag;
}
};

int len;
Complex wn[MAXN],wn_anti[MAXN];

void FFT(Complex y[],int op){
for(int i=1,j=len>>1,k; i<len-1; ++i){
if(i<j) swap(y[i],y[j]);
k=len>>1;
while(j>=k){
j-=k;
k>>=1;
}
if(j<k) j+=k;
}
for(int h=2; h<=len; h<<=1){
Complex Wn=(op==1?wn[h]:wn_anti[h]);
for(int i=0; i<len; i+=h){
Complex W(1,0);
for(int j=i; j<i+(h>>1); ++j){
Complex u=y[j],t=W*y[j+(h>>1)];
y[j]=u+t;
y[j+(h>>1)]=u-t;
W=W*Wn;
}
}
}
if(op==-1){
for(int i=0; i<len; ++i) y[i].real/=len;
}
}
void Convolution(Complex A[],Complex B[],int n){
for(len=1; len<(n<<1); len<<=1);
for(int i=n; i<len; ++i){
A[i].setValue();
B[i].setValue();
}

FFT(A,1); FFT(B,1);
for(int i=0; i<len; ++i){
A[i]=A[i]*B[i];
}
FFT(A,-1);
}

int a[111111],d[111111];
Complex A[MAXN],B[MAXN];

void cdq(int l,int r){
if(l==r){
d[l]+=a[l];
d[l]%=313;
return;
}
int mid=l+r>>1;
cdq(l,mid);

for(int i=l; i<=mid; ++i) A[i-l].setValue(d[i]);
for(int i=0; i<=r-l; ++i) B[i].setValue(a[i]);
for(int i=mid-l+1; i<=r-l; ++i) A[i].setValue();
Convolution(A,B,r-l+1);
for(int i=mid+1; i<=r; ++i){
d[i]+=((long long)(A[i-l].real+0.5))%313;
d[i]%=313;
}

cdq(mid+1,r);
}

int main(){
for(int i=0; i<MAXN; ++i){
wn[i].setValue(cos(2.0*PI/i),sin(2.0*PI/i));
wn_anti[i].setValue(wn[i].real,-wn[i].imag);
}
int n;
while(~scanf("%d",&n) && n){
for(int i=1; i<=n; ++i){
scanf("%d",a+i);
a[i]%=313;
}
memset(d,0,sizeof(d));
cdq(1,n);
printf("%d\n",d
);
}
return 0;
}
内容来自用户分享和网络整理,不保证内容的准确性,如有侵权内容,可联系管理员处理 点击这里给我发消息
标签: