• CF623E Transforming Sequence


    嘟嘟嘟


    我认为这题是黑题的原因是质数不是(998244353),所以得用三模NTT或是拆系数FFT。我抄了一个拆系数FFT的板子,但现在暂时还是不很懂。
    但这不影响解题思路。


    首先(n> K)无解。(完全搞不懂(n)那么大干啥)


    我们令(dp[i][j])表示第(i)个数有(j)(1)时的方案数。我们先不考虑(1)的位置,这样答案就是(sum _ {i = 1} ^ {K} dp[n][i] * inom{n}{i})。转移也很显然,因为是(or)操作,所以原来是(1)的位置填不填(1)都行,即(dp[i][j] = sum _ {k = 0} ^ {j}dp[i - 1][k] * inom{j}{k} * 2 ^ k)。把组合数拆开后可以用FFT加速。那么现在可以在(O(k ^ 2logk))的时间内解决了。


    现在的瓶颈在于只能一位一位的dp,得循环(n)次。我们想办法优化:考虑两个dp方程的合并,就能得出

    [dp[x + y][i] = sum _{j = 0} ^ {i}dp[x][j] * dp[y][i - j] *inom{i}{j} *2 ^{y * j} ]

    然后把它整理一下:

    [frac{dp[x + y][i]}{i!} = sum _ {j = 0} ^ {i} frac{dp[x][j] * (2 ^ y) ^ j}{j!} * frac{dp[y][i - j]}{(i - j)!} ]

    于是愉快的多项式快速幂就可以啦!


    代码里的过程量都是(frac{dp[i][j]}{i!}),到最后再乘上(i!)即可。

    #include<cstdio>
    #include<iostream>
    #include<cmath>
    #include<algorithm>
    #include<cstring>
    #include<cstdlib>
    #include<cctype>
    #include<vector>
    #include<queue>
    #include<assert.h>
    #include<ctime>
    using namespace std;
    #define enter puts("")
    #define space putchar(' ')
    #define Mem(a, x) memset(a, x, sizeof(a))
    #define In inline
    #define forE(i, x, y) for(int i = head[x], y; (y = e[i].to) && ~i; i = e[i].nxt)
    typedef long long ll;
    typedef long double db;
    const int INF = 0x3f3f3f3f;
    const db eps = 1e-8;
    const int maxn = 3e5 + 5;
    const ll mod = 1e9 + 7;
    const db PI = acos(-1);
    In ll read()
    {
    	ll ans = 0;
    	char ch = getchar(), las = ' ';
    	while(!isdigit(ch)) las = ch, ch = getchar();
    	while(isdigit(ch)) ans = (ans << 1) + (ans << 3) + ch - '0', ch = getchar();
    	if(las == '-') ans = -ans;
    	return ans;
    }
    In void write(ll x)
    {
    	if(x < 0) x = -x, putchar('-');
    	if(x >= 10) write(x / 10);
    	putchar(x % 10 + '0');
    }
    In void MYFILE()
    {
    #ifndef mrclr
    	freopen("ha.in", "r", stdin);
    	freopen("hia.out", "w", stdout);
    #endif
    }
    
    ll n;
    int m;
    ll fac[maxn], inv[maxn], p2[maxn];
    
    In ll inc(ll a, ll b) {return a + b < mod ? a + b : a + b - mod;}
    In ll C(int n, int m)
    {
    	if(m > n) return 0;
    	return fac[n] * inv[m] % mod * inv[n - m] % mod;
    }
    In ll quickpow(ll a, ll b)
    {
    	ll ret = 1;
    	for(; b; b >>= 1, a = a * a % mod)
    		if(b & 1) ret = ret * a % mod;
    	return ret;
    }
    
    In void init()
    {
    	fac[0] = inv[0] = p2[0] = 1;
    	for(int i = 1; i < maxn; ++i) fac[i] = fac[i - 1] * i % mod;
    	inv[maxn - 1] = quickpow(fac[maxn - 1], mod - 2);
    	for(int i = maxn - 2; i; --i) inv[i] = inv[i + 1] * (i + 1) % mod;
    	for(int i = 1; i < maxn; ++i) p2[i] = (p2[i - 1] * 2) % mod;
    }
    
    int len = 1, lim = 0, rev[maxn << 2];
    struct Comp
    {
    	db x, y;
    	In Comp operator + (const Comp& oth)const
    	{
    		return (Comp){x + oth.x, y + oth.y};
    	}
    	In Comp operator - (const Comp& oth)const
    	{
    		return (Comp){x - oth.x, y - oth.y};
    	}
    	In Comp operator * (const Comp& oth)const
    	{
    		return (Comp){x * oth.x - y * oth.y, x * oth.y + y * oth.x};
    	}
    	friend In void swap(Comp& A, Comp& B)
    	{
    		swap(A.x, B.x), swap(A.y, B.y);
    	}
    	friend In Comp operator ! (Comp a)
    	{
    		return (Comp){a.x, -a.y};
    	}
    }A[maxn << 2], B[maxn << 2];
    Comp dftA[maxn << 2], dftB[maxn << 2], dftC[maxn << 2], dftD[maxn << 2];
    In void fft(Comp* a, int len, int flg)
    {
    	for(int i = 0; i < len; ++i) if(i < rev[i]) swap(a[i], a[rev[i]]);
    	for(int i = 1; i < len; i <<= 1)
    	{
    		Comp omg = (Comp){cos(PI / i), sin(PI / i) * flg};
    		for(int j = 0; j < len; j += (i << 1))
    		{
    			Comp o = (Comp){1, 0};
    			for(int k = 0; k < i; ++k, o = o * omg)
    			{
    				Comp tp1 = a[j + k], tp2 = o * a[j + k + i];
    				a[j + k] = tp1 + tp2, a[j + k + i] = tp1 - tp2;
    			}
    		}
    	}
    }
    const ll NUM = 32767;
    In void FFT(ll* a, ll* b, ll* c, int n)
    {
    	for(int i = 0; i < len; ++i)
    	{
    		ll tp1 = i <= n ? a[i] : 0;
    		ll tp2 = i <= n ? b[i] : 0;
    		A[i] = (Comp){tp1 & NUM, tp1 >> 15};
    		B[i] = (Comp){tp2 & NUM, tp2 >> 15};
    	}
    	fft(A, len, 1), fft(B, len, 1);
    	for(int i = 0; i < len; ++i)
    	{
    		int j = (len - i) & (len - 1);
    		Comp da = (A[i] + (!A[j])) * (Comp){0.5, 0};
    		Comp db = (A[i] - (!A[j])) * (Comp){0, -0.5};
    		Comp dc = (B[i] + (!B[j])) * (Comp){0.5, 0};
    		Comp dd = (B[i] - (!B[j])) * (Comp){0, -0.5};
    		dftA[i] = da * dc, dftB[i] = da * dd;
    		dftC[i] = db * dc, dftD[i] = db * dd;
    	}
    	for(int i = 0; i < len; ++i)
    	{
    		A[i] = dftA[i] + dftB[i] * (Comp){0, 1};
    		B[i] = dftC[i] + dftD[i] * (Comp){0, 1};
    	}
    	fft(A, len, -1), fft(B, len, -1);
    	for(int i = 0; i <= n; ++i)
    	{
    		ll da = (ll)(A[i].x / len + 0.5) % mod;
    		ll db = (ll)(A[i].y / len + 0.5) % mod;
    		ll dc = (ll)(B[i].x / len + 0.5) % mod;
    		ll dd = (ll)(B[i].y / len + 0.5) % mod;
    		c[i] = inc(inc(da, ((db + dc) << 15) % mod), (dd << 30) % mod);
    	}
    }
    
    ll dp[maxn], f[maxn], ta[maxn], LEN = 1;
    In void mul(ll* a, ll* b)
    {
    	ll tp = 1;
    	for(int i = 0; i <= m; ++i) 
    	{
    		ta[i] = a[i] * tp  % mod;
    		tp = tp * p2[LEN] % mod;
    	}
    	FFT(ta, b, ta, m);
    	for(int i = 0; i <= m; ++i) a[i] = ta[i] % mod;
    	
    }
    
    In ll QuickPow(ll n)
    {
    	for(int i = 1; i <= m; ++i) dp[i] = inv[i];
    	f[0] = 1;
    	LEN = 1;
    	for(; n; n >>= 1, mul(dp, dp), LEN <<= 1)
    		if(n & 1) mul(f, dp);
    	ll ret = 0;
    	for(int i = 1; i <= m; ++i) ret = inc(ret, f[i] * fac[i] % mod* C(m, i) % mod);
    	return ret;
    }
    
    int main()
    {
    //	MYFILE();
    	init();
    	n = read(), m = read();
    	if(n > m) {puts("0"); return 0;}
    	while(len <= (m << 1)) len <<= 1, ++lim;
    	for(int i = 0; i < len; ++i) rev[i] = (rev[i >> 1] >> 1) | ((i & 1) << (lim - 1));
    	write(QuickPow(n)), enter;
    	return 0;
    }
    

    这题昨晚调了半天还是WA,回宿舍后,**多项式大师** $S color{red}{jk}$看了一眼后说道,你把double改成long double,就过了。然后还真就过了,神啊!
  • 相关阅读:
    携程 去呼呼 Odoo SSO 单点登录
    Spring Boot 2.2 正式发布,大幅性能提升 + Java 13 支持 & Spring Data Moore M2
    全球顶尖AI技术 微软CRM X 销售易CRM Gartner魔力象限
    职场的真相——七句话
    东软 UniEAP SaCa DataViz & Report
    UML建模——活动图(Activity Diagram)
    领域驱动设计(DDD)部分核心概念
    阿里云RPA4.0背后的黑科技
    Open Source In-Memory Computing Platform
    Mac OS X ifconfig命令解释
  • 原文地址:https://www.cnblogs.com/mrclr/p/11118122.html
Copyright © 2020-2023  润新知