给出长度分别为 $ 1 \sim n $ 的珠子,长度为i的珠子有$a[i]$种,每种珠子有无限个,问用这些珠子串成长度为n的链有多少种方案。
链接
题解
( 多么好的一个模板题!业界良心!
令dp[i]表示用这些珠子串成长度为i的链的方案数,并令dp[0]=1,轻松得到转移方程
为$dp[i]=dp[i-1]a[1]+dp[i-2]a[2]+ … +dp[1]a[i-1]+dp[0]a[i]$
即 $dp[i]=\sum_{j=0}^{i-1}dp[j]*a[i-j]$
解法一
对于这个式子一个显然的套路就是CDQ分治+FFT
现在考虑用cdq分治处理区间$[l,r]$,按照常规,$mid=(l+r)/2,$我们可以先递归处理$[l,mid]$区间,
然后将$[l,mid]$区间对$[mid+1,r]$区间的影响累加上去,
现在假设$cdq(l,mid)$函数已经求出了$dp[l],dp[l+1]…dp[mid]$的值,
我们考虑如何将其对后半段区间的影响累加上去:
设$g[i]$表示$dp[l],dp[l+1]…dp[mid]$对$dp[i]$的影响,
显然由最开始的DP转移式子,我们可以得出
$g[i]=dp[l]a[i-l]+dp[l+1]a[i-l-1]+…=\sum_{j=l}^{mid}dp[j]*a[i-j]$
那么对于区间$[l,mid]$和$[mid+1,r]$,
$dp[l] \ \ dp[l+1] \ \ … \ \ dp[mid] $
$a[1] \ \ a[2] \ \ a[3] \ \ … \ \ a[mid+1-l]$
把以上两个式子进行卷积即可,
然后把对应的影响累加到$dp[mid+1],dp[mid+2]…dp[r]$上,
再用cdq处理右半部分区间即可。
注意不要忘了l==r时的终止条件。
解法二
对 $f_n=\sum_{i=0}^{n-1}f_i*a_{n - i}$ 考虑生成函数。
设 $A(x) = a_0x^0 + a_1x^1 + a_2x^2 + …$
设 $F(x) = f_0x^0 + f_1x^1 + f_2x^2 + …$
显然
$$
\begin{align}
A(x) * F(x) &= 0 + a_1x^1 + (a_1f_1 + a_2f_0)x^2 + …\
&= F(x) - 1
\end{align}
$$
即
$$F(x) = \dfrac{1}{1 - A(x)} $$
那么直接用FFT多项式求逆即可,由于模数很小仅为 $313$
理论上可以用double直接怼过去(我没试过啊口胡的 CDQ分治+FFT就是直接double怼过去的)
比较稳妥的做法是用NTT进行多项式求逆,但是并不是一个很舒服的做法
因为 $313$ 并不是一个特殊的素数,常规做法需要CRT进行合并
但是我们现在有了新的黑科技!叫做拆系数FFT(也称MTT),可以处理任意模数FFT问题
同时这个博客中还提到了一种FFT卡常数技巧……
代码
解法一
CDQ分治 + FFT
/*
* Filename: hdu5730.cpp
* Created: Tuesday, October 17, 2017 12:57:58 PM
* Author: crazyX
* More:
*
*/
#include <bits/stdc++.h>
#define mp make_pair
#define pb push_back
#define fi first
#define se second
#define SZ(x) ((int) (x).size())
#define all(x) (x).begin(), (x).end()
#define sqr(x) ((x) * (x))
#define clr(a,b) (memset(a,b,sizeof(a)))
#define y0 y3487465
#define y1 y8687969
#define fastio std::ios::sync_with_stdio(false)
using namespace std;
typedef long long ll;
typedef double DB;
const int maxn = 1e5 + 7;
const int maxLen = 18, maxm = 1 << maxLen | 1;
const ll maxv = 1e10 + 6; // 1e14, 1e15
const DB pi = acos(-1.0); // double is enough
ll mod = 313, nlim, sp, msk;
#define _ %mod
#define __ %=mod
ll qpow(ll x, ll p) {
ll ret = 1;
while (p) {
if (p & 1) (ret *= x) __;
(x *= x) __;
p >>= 1;
}
return ret;
}
namespace FFT{
struct cp {
DB r, i;
cp() {}
cp(DB r, DB i) : r(r), i(i) {}
cp operator + (cp const &t) const { return cp(r + t.r, i + t.i); }
cp operator - (cp const &t) const { return cp(r - t.r, i - t.i); }
cp operator * (cp const &t) const { return cp(r * t.r - i * t.i, r * t.i + i * t.r); }
cp conj() const { return cp(r, -i); }
} w[maxm], wInv[maxm];
void init() {
for(int i = 0, ilim = 1 << maxLen; i < ilim; ++i) {
int j = i, k = ilim >> 1; // 2 pi / ilim
for( ; !(j & 1) && !(k & 1); j >>= 1, k >>= 1);
w[i] = cp(cos(pi / k * j), sin(pi / k * j));
wInv[i] = w[i].conj();
}
nlim = std::min(maxv / (mod - 1) / (mod - 1), maxn - 1LL);
for(sp = 1; 1 << (sp << 1) < mod; ++sp);
msk = (1 << sp) - 1;
}
void FFT(int n, cp a[], int flag) {
static int bitLen = 0, bitRev[maxm] = {};
if(n != (1 << bitLen)) {
for(bitLen = 0; 1 << bitLen < n; ++bitLen);
for(int i = 1; i < n; ++i)
bitRev[i] = (bitRev[i >> 1] >> 1) | ((i & 1) << (bitLen - 1));
}
for(int i = 0; i < n; ++i)
if(i < bitRev[i])
std::swap(a[i], a[bitRev[i]]);
for(int i = 1, d = 1; d < n; ++i, d <<= 1)
for(int j = 0; j < n; j += d << 1)
for(int k = 0; k < d; ++k) {
cp &AL = a[j + k], &AH = a[j + k + d];
cp TP = w[k << (maxLen - i)] * AH;
AH = AL - TP, AL = AL + TP;
}
if(flag != -1)
return;
std::reverse(a + 1, a + n);
for(int i = 0; i < n; ++i) {
a[i].r /= n;
a[i].i /= n;
}
}
void polyMul(int a[], int aLen, int b[], int bLen, int c[]) { // c not in {a, b}
static cp A[maxm], B[maxm], C[maxm], D[maxm];
int len, cLen = aLen + bLen - 1; // optional: parameter
for(len = 1; len < aLen + bLen - 1; len <<= 1);
if(std::min(aLen, bLen) <= nlim) {
for(int i = 0; i < len; ++i)
A[i] = cp(i < aLen ? a[i] : 0, i < bLen ? b[i] : 0);
FFT(len, A, 1);
cp tr(0, -0.25);
for(int i = 0, j; i < len; ++i)
j = (len - i) & (len - 1), B[i] = (A[i] * A[i] - (A[j] * A[j]).conj()) * tr;
FFT(len, B, -1);
for(int i = 0; i < cLen; ++i) c[i] = (ll)(B[i].r + 0.5) % mod;
return;
} // if min(aLen, bLen) * mod <= maxv
for(int i = 0; i < len; ++i) {
A[i] = i < aLen ? cp(a[i] & msk, a[i] >> sp) : cp(0, 0);
B[i] = i < bLen ? cp(b[i] & msk, b[i] >> sp) : cp(0, 0);
}
FFT(len, A, 1), FFT(len, B, 1);
cp trL(0.5, 0), trH(0, -0.5), tr(0, 1);
for(int i = 0, j; i < len; ++i) {
j = (len - i) & (len - 1);
cp AL = (A[i] + A[j].conj()) * trL;
cp AH = (A[i] - A[j].conj()) * trH;
cp BL = (B[i] + B[j].conj()) * trL;
cp BH = (B[i] - B[j].conj()) * trH;
C[i] = AL * (BL + BH * tr);
D[i] = AH * (BL + BH * tr);
}
FFT(len, C, -1), FFT(len, D, -1);
for(int i = 0; i < cLen; ++i) {
int v11 = (ll)(C[i].r + 0.5) % mod, v12 = (ll)(C[i].i + 0.5) % mod;
int v21 = (ll)(D[i].r + 0.5) % mod, v22 = (ll)(D[i].i + 0.5) % mod;
c[i] = (((((ll)v22 << sp) + v12 + v21) << sp) + v11) % mod;
}
}
};
int n, ar[maxn], dp[maxn], a[maxn], b[maxn], c[maxm];
void cdq(int l,int r){
if(l==r){
(dp[l]+=ar[l])__;
return;
}
int mid=(l+r)>>1;
cdq(l,mid);
for (int i = l; i <= mid; i += 1)
b[i-l]=dp[i];
for (int i = 0; i <= r-l+1; i += 1)
a[i]=ar[i];
FFT::polyMul(a,r-l+1+1,b,mid-l+1,c);
for (int i = mid+1; i <= r; i += 1)
(dp[i]+=ll(c[i-l-1]+0.5))__;
cdq(mid+1,r);
}
int main()
{
#ifdef AC
freopen("data.in", "r", stdin);
//freopen("data.out", "w", stdout);
#endif
FFT::init();
a[0] = 1;
while (scanf("%d", &n) && n) {
for (int i = 0; i < n; i += 1) scanf("%d", ar + i), ar[i] __, dp[i] = 0;
cdq(0, n - 1);
printf("%d\n", dp[n - 1]);
}
return 0;
}
解法二
多项式求逆
/*
* Filename: hdu5730.cpp
* Created: Tuesday, October 17, 2017 12:57:58 PM
* Author: crazyX
* More:
*
*/
#include <bits/stdc++.h>
#define mp make_pair
#define pb push_back
#define fi first
#define se second
#define SZ(x) ((int) (x).size())
#define all(x) (x).begin(), (x).end()
#define sqr(x) ((x) * (x))
#define clr(a,b) (memset(a,b,sizeof(a)))
#define y0 y3487465
#define y1 y8687969
#define fastio std::ios::sync_with_stdio(false)
using namespace std;
typedef long long ll;
typedef double DB;
const int maxn = 1e5 + 7;
const int maxLen = 18, maxm = 1 << maxLen | 1;
const ll maxv = 1e10 + 6; // 1e14, 1e15
const DB pi = acos(-1.0); // double is enough
ll mod = 313, nlim, sp, msk;
#define _ %mod
#define __ %=mod
ll qpow(ll x, ll p) {
ll ret = 1;
while (p) {
if (p & 1) (ret *= x) __;
(x *= x) __;
p >>= 1;
}
return ret;
}
namespace FFT{
struct cp {
DB r, i;
cp() {}
cp(DB r, DB i) : r(r), i(i) {}
cp operator + (cp const &t) const { return cp(r + t.r, i + t.i); }
cp operator - (cp const &t) const { return cp(r - t.r, i - t.i); }
cp operator * (cp const &t) const { return cp(r * t.r - i * t.i, r * t.i + i * t.r); }
cp conj() const { return cp(r, -i); }
} w[maxm], wInv[maxm];
void init() {
for(int i = 0, ilim = 1 << maxLen; i < ilim; ++i) {
int j = i, k = ilim >> 1; // 2 pi / ilim
for( ; !(j & 1) && !(k & 1); j >>= 1, k >>= 1);
w[i] = cp(cos(pi / k * j), sin(pi / k * j));
wInv[i] = w[i].conj();
}
nlim = std::min(maxv / (mod - 1) / (mod - 1), maxn - 1LL);
for(sp = 1; 1 << (sp << 1) < mod; ++sp);
msk = (1 << sp) - 1;
}
void FFT(int n, cp a[], int flag) {
static int bitLen = 0, bitRev[maxm] = {};
if(n != (1 << bitLen)) {
for(bitLen = 0; 1 << bitLen < n; ++bitLen);
for(int i = 1; i < n; ++i)
bitRev[i] = (bitRev[i >> 1] >> 1) | ((i & 1) << (bitLen - 1));
}
for(int i = 0; i < n; ++i)
if(i < bitRev[i])
std::swap(a[i], a[bitRev[i]]);
for(int i = 1, d = 1; d < n; ++i, d <<= 1)
for(int j = 0; j < n; j += d << 1)
for(int k = 0; k < d; ++k) {
cp &AL = a[j + k], &AH = a[j + k + d];
cp TP = w[k << (maxLen - i)] * AH;
AH = AL - TP, AL = AL + TP;
}
if(flag != -1)
return;
std::reverse(a + 1, a + n);
for(int i = 0; i < n; ++i) {
a[i].r /= n;
a[i].i /= n;
}
}
void polyMul(int a[], int aLen, int b[], int bLen, int c[]) { // c not in {a, b}
static cp A[maxm], B[maxm], C[maxm], D[maxm];
int len, cLen = aLen + bLen - 1; // optional: parameter
for(len = 1; len < aLen + bLen - 1; len <<= 1);
if(std::min(aLen, bLen) <= nlim) {
for(int i = 0; i < len; ++i)
A[i] = cp(i < aLen ? a[i] : 0, i < bLen ? b[i] : 0);
FFT(len, A, 1);
cp tr(0, -0.25);
for(int i = 0, j; i < len; ++i)
j = (len - i) & (len - 1), B[i] = (A[i] * A[i] - (A[j] * A[j]).conj()) * tr;
FFT(len, B, -1);
for(int i = 0; i < cLen; ++i) c[i] = (ll)(B[i].r + 0.5) % mod;
return;
} // if min(aLen, bLen) * mod <= maxv
for(int i = 0; i < len; ++i) {
A[i] = i < aLen ? cp(a[i] & msk, a[i] >> sp) : cp(0, 0);
B[i] = i < bLen ? cp(b[i] & msk, b[i] >> sp) : cp(0, 0);
}
FFT(len, A, 1), FFT(len, B, 1);
cp trL(0.5, 0), trH(0, -0.5), tr(0, 1);
for(int i = 0, j; i < len; ++i) {
j = (len - i) & (len - 1);
cp AL = (A[i] + A[j].conj()) * trL;
cp AH = (A[i] - A[j].conj()) * trH;
cp BL = (B[i] + B[j].conj()) * trL;
cp BH = (B[i] - B[j].conj()) * trH;
C[i] = AL * (BL + BH * tr);
D[i] = AH * (BL + BH * tr);
}
FFT(len, C, -1), FFT(len, D, -1);
for(int i = 0; i < cLen; ++i) {
int v11 = (ll)(C[i].r + 0.5) % mod, v12 = (ll)(C[i].i + 0.5) % mod;
int v21 = (ll)(D[i].r + 0.5) % mod, v22 = (ll)(D[i].i + 0.5) % mod;
c[i] = (((((ll)v22 << sp) + v12 + v21) << sp) + v11) % mod;
}
}
int c[maxm], tmp[maxm];
// y should clear to 0
void polyInv(int x[], int y[], int deg) {
if (deg == 1) {
y[0] = qpow(x[0], mod - 2);
return;
}
polyInv(x, y, (deg + 1) >> 1);
copy(x, x + deg, tmp);
int p = ((deg + 1) >> 1) + deg - 1;
polyMul(y, (deg + 1) >> 1, tmp, deg, c);
for (int i = 0; i < p; i += 1) c[i] = (- c[i] + mod) _;
(c[0] += 2) __;
polyMul(y, (deg + 1) >> 1, c, deg, tmp);
copy(tmp, tmp + deg, y);
}
};
int n, a[maxn], b[maxn];
int main()
{
#ifdef AC
freopen("data.in", "r", stdin);
//freopen("data.out", "w", stdout);
#endif
FFT::init();
a[0] = 1;
while (scanf("%d", &n) && n) {
for (int i = 0; i <= n; i += 1) b[i] = 0;
for (int i = 1; i <= n; i += 1) scanf("%d", a + i), a[i] = mod - a[i] _;
FFT::polyInv(a, b, n + 1);
printf("%d\n", b[n]);
}
return 0;
}