HDU-6050 Funny Function

2017 Multi-University Training Contest - Team 2 - 1006

HDU-6050 Funny Function

---

题意：

给出n, m
n代表 $F_{i,j}$  等于上一行从j开始往后n个元素的和
然后求第m行第一个元素 $F_{m,1}$ 的值

---

思路：

先证

$F_{i, j} = F_{i, j - 1} + 2 * F_{i, j - 2}$

这里用数学归纳法来证明

取 i = 2 ，n = 1

则$F_{2, j} = F_{1, j}$ 可以推出 $F_{2, j} = F_{2, j - 1} + 2 * F_{2, j - 2}$

假设 n = x 时 $F_{2, j} = F_{2, j - 1} + 2 * F_{2, j - 2}$ 成立

即 $\sum_{k = j}^{j + 2+ x - 1} F_{1, k}= \sum_{k = j + 1}^{j + 1 + x - 1} F_{1, k}+ 2 * \sum_{k = j}^{j + x - 1} F_{1, k}$    - - - - - -    ① 成立

当 n = x + 1 时

$F_{2, j + 2} = \sum_{k = j + 2}^{j + 2 + x} F_{1, k} = \sum_{k = j + 2}^{j + 2 + x - 1} F_{1, k}+ F_{1, j + 2 + x}$

$F_{2, j + 1} = \sum_{k = j + 1}^{j + 1 + x} F_{1, k} = \sum_{k = j + 1}^{j + 1 + x - 1} F_{1, k}+ F_{1, j + 1 + x}$

$F_{2, j} = \sum_{k = j}^{j + x} F_{1, k} = \sum_{k = j}^{j + x - 1} F_{1, k}+ F_{1, j + x}$

由题目中 $F_{1， j} = F_{1， j - 1} + 2 * F_{1, j - 2}$ 得出 $F_{1， j + 2} = F_{1， j + 1} + 2 * F_{1, j}$

又因为 ① 所以我们可以得出 $F_{2， j} = F_{2， j - 1} + 2 * F_{2, j - 2}$ 在 $n \ge 1$时成立

然后以此类推 i = 3, 4, 5, 6 ······ 时也成立

证毕

再证

$\sum_{k = 1}^{j} F_{i, k} = F_{i, j + 1} - F_{i, 1}$ ( j 为偶数时)

$\sum_{k = 1}^{j} F_{i, k} = F_{i, j + 1} + F_{i, 1} - F_{i, 2}$ ( j 为奇数时)

由上一步的公式 $F_{i, j} = F_{i, j - 1} + 2 * F_{i, j - 2}$ 得

---

当 j 为偶数时

$F_{i, j + 1} = F_{i, j} + 2 * F_{i, j - 1}$

$F_{i, j - 1} = F_{i, j - 2} + 2 * F_{i, j - 3}$

$F_{i, j - 3} = F_{i, j - 4} + 2 * F_{i, j - 5}$

······

$F_{i, 3} = F_{i, 2} + 2 * F_{i, 1}$

我们把等号左右分别加起来得到

$F_{i, j + 1} = F_{i, j} + F_{i, j - 1} + F_{i, j - 2} + ··· + F_{i, 2} + 2 * F_{i, 1}$

           $= F_{i, j} + F_{i, j - 1} + F_{i, j - 2} + ··· + F_{i, 2} + F_{i, 1} + F_{i, 1}$

           $= \sum_{k = 1}^{j} F_{i, j} + F_{i, 1}$

移项得 $\sum_{k = 1}^{j} F_{i, k} = F_{i, j + 1} - F_{i, 1}$ ( j 为偶数时)

---

当 j 为奇数时

$F_{i, j + 1} = F_{i, j} + 2 * F_{i, j - 1}$

$F_{i, j - 1} = F_{i, j - 2} + 2 * F_{i, j - 3}$

$F_{i, j - 3} = F_{i, j - 4} + 2 * F_{i, j - 5}$

······

$F_{i, 4} = F_{i, 3} + 2 * F_{i, 2}$

把等号左右分别加起来得

$F_{i, j + 1} = F_{i, j} + F_{i, j - 1} + F_{i, j - 2} + ··· + F_{i, 2} + 2 * F_{i, 2}$

等号两边分别加上 $F_{i, 1}$ 得

$F_{i, j + 1} + F_{i, 1}= F_{i, j} + F_{i, j - 1} + F_{i, j - 2} + ··· + F_{i, 2} + F_{i, 2} + F_{i, 1} + F_{i, 2}$

$F_{i, j + 1} + F_{i, 1}= \sum_{k = 1}^{j} F_{i, k} + F_{i, 2}$

移项得 $\sum_{k = 1}^{j} F_{i, k} = F_{i, j + 1} + F_{i, 1} - F_{i, 2}$ ( j 为奇数时)

证毕

然后是

$(F_{m, 1}, F_{m, 2}) = (F_{1, 1}, F_{1, 2})(A^{n} - B_{0})^{m - 1}$ ( n 为偶数时)

$(F_{m, 1}, F_{m, 2}) = (F_{1, 1}, F_{1, 2})(A^{n} - B_{1})^{m - 1}$ ( n 为奇数时)

由第一步的递推公式 $F_{i, j} = F_{i, j - 1} + 2 * F_{i, j - 2}$ 来构造矩阵

$(F_{i, j - 1}, F_{i, j}) = (F_{i, j - 2}, F_{i, j - 1}) * \begin{pmatrix} 0 & 2 \\ 1 & 1 \end{pmatrix}$

令 $A = \begin{pmatrix} 0 & 2 \\ 1 & 1 \end{pmatrix}$

则 $(F_{i, j}, F_{i, j + 1}) = (F_{i, j - 1}, F_{i, j}) * A = (F_{i, 1}, F_{i, 2}) * A^{j - 1}$    - - - - - -    ①

---

当 n 为偶数时

由题目所给公式可得

$F_{i, 1} = \sum_{k = 1}^{n} F_{i - 1, k}$         由第二步公式得

        $= F_{i - 1, n + 1} - F_{i - 1, 1}$

$F_{i, 2} = \sum_{k = 1}^{n + 1} F_{i - 1, k} - F_{i - 1, 1}$        由第二步公式得

        $= F_{i - 1, n + 2} + F_{i - 1, 1} - F_{i - 1, 2} - F_{i - 1, 1}$

        $= F_{i - 1, n + 2} - F_{i - 1, 2}$

则可以得出

$(F_{i, 1}, F_{i, 2}) = (F_{i - 1, n + 1} - F_{i - 1, 1} , F_{i - 1, n + 2} - F_{i - 1, 2})$

                   $= (F_{i - 1, n + 1} , F_{i - 1, n + 2}) - (F_{i - 1, 1} , F_{i - 1, 2})$         令①中 j = n + 1 带入得

                   $= (F_{i - 1, 1} , F_{i - 1, 2}) * A^{n} - (F_{i - 1, 1} , F_{i - 1, 2}) * \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}$

令 $B_{0} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}$ 则可以得到

$(F_{i, 1} , F_{i, 2}) = (F_{i - 1, 1} , F_{i - 1, 2}) * (A^{n} - B_{0})$

而这又是一个递推关系，我们就可以得到

$(F_{m, 1} , F_{m , 2}) = (F_{1, 1} , F_{1, 2}) * (A^{n} - B_{0})^{m - 1}$

---

当 n 为奇数时

$F_{i, 1} = \sum_{k = 1}^{n} F_{i - 1, k}$        由第二步公式得

        $= F_{i - 1, n + 1} + F_{i - 1, 1} - F_{i - 1, 2}$

$F_{i, 2} = \sum_{k = 1}^{n + 1} F_{i - 1, k} - F_{i - 1, 1}$        由第二步公式得

        $= F_{i - 1, n + 2} - F_{i - 1, 1} - F_{i - 1, 1}$

        $= F_{i - 1, n + 2} - 2 * F_{i - 1, 1}$

则可以得出

$(F_{i, 1}, F_{i, 2}) = (F_{i - 1, n + 1} + F_{i - 1, 1} - F_{i - 1, 2} , F_{i - 1, n + 2} - 2 * F_{i - 1, 1} )$

                   $= (F_{i - 1, n + 1} , F_{i - 1 , n + 2}) - (-F_{i - 1, 1} + F_{i - 1, 2} , 2 * F_{i - 1, 1})$         令①中 j = n + 1 带入得

                   $= (F_{i - 1, 1} , F_{i - 1, 2}) * A^{n} - (F_{i - 1, 1} , F_{i - 1, 2}) * \begin{pmatrix} -1 & 2 \\ 1 & 0 \end{pmatrix}$

令 $B_{1} = \begin{pmatrix} -1 & 2 \\ 1 & 0 \end{pmatrix}$ 则可以得到

$(F_{i, 1} , F_{i, 2}) = (F_{i - 1, 1} , F_{i - 1, 2}) * (A^{n} - B_{1})$

而这又是一个递推关系，我们就可以得到

$(F_{m, 1} , F_{m , 2}) = (F_{1, 1} , F_{1, 2}) * (A^{n} - B_{1})^{m - 1}$

综上我们得出了两个公式

$(F_{m, 1}, F_{m, 2}) = (F_{1, 1}, F_{1, 2})(A^{n} - B_{0})^{m - 1}$ ( n 为偶数时)

$(F_{m, 1}, F_{m, 2}) = (F_{1, 1}, F_{1, 2})(A^{n} - B_{1})^{m - 1}$ ( n 为奇数时)

及公式中的$A, B_{0}, B_{1}$矩阵

$A = \begin{pmatrix} 0 & 2 \\ 1 & 1 \end{pmatrix}$      $B_{0} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}$      $B_{1} = \begin{pmatrix} -1 & 2 \\ 1 & 0 \end{pmatrix}$

然后就可以愉快的使用矩阵快速幂求解了

---

代码：

#include <cstdio>
#include <iostream>
#include <cstring>
using namespace std;

const int mod = 1e9 + 7;
struct S
{
    long long a[2][2];
};
S cheng(S a, S b)  // 矩阵乘法
{
    S temp;
    memset(temp.a, 0, sizeof(temp.a));
    for(int i = 0; i < 2; ++i)
    {
        for(int j = 0; j < 2; ++j)
        {
            for(int k = 0; k  < 2; ++k)
            {
                temp.a[i][j] += a.a[i][k] * b.a[k][j] % mod;
                temp.a[i][j] %= mod;
            }
        }
    }
    return temp;
}

int main()
{
    S a, b[2], f, temp;
    int t;
    long long n, m, n1;
    while(~scanf("%d", &t))
    {
        while(t--)
        {
            memset(a.a, 0, sizeof(a.a));  // 初始化 A B0 B1 矩阵
            memset(b[0].a, 0, sizeof(b[0].a));
            memset(b[1].a, 0, sizeof(b[1].a));
            memset(f.a, 0, sizeof(f.a));
            memset(temp.a, 0, sizeof(temp.a));
            a.a[0][1] = b[1].a[0][1] = 2;
            a.a[1][0] = a.a[1][1] = b[0].a[0][0] = b[0].a[1][1] = b[1].a[1][0]  = f.a[0][0] = f.a[0][1] = 1;
            b[1].a[0][0] = -1;
            temp.a[0][0] = temp.a[1][1] = 1;

            scanf("%lld%lld", &n, &m);
            n1 = n;
            while(n)  // temp保存A^n的值
            {
                if(n & 1)
                {
                    temp = cheng(temp, a);
                }
                n >>= 1;
                a = cheng(a, a);
            }

            for(int i = 0; i < 2; ++i)  // A 保存 A^n - B 的值
            {
                for(int j = 0; j < 2; ++j)
                {
                    a.a[i][j] = temp.a[i][j] - b[n1 % 2].a[i][j];
                }
            }

            memset(temp.a, 0, sizeof(temp.a));
            temp.a[0][0] = temp.a[1][1] = 1;
            --m;
            while(m)  // 求(A^n - B)^(m - 1)保存在temp中
            {
                if(m & 1)
                {
                    temp = cheng(temp, a);
                }
                m >>= 1;
                a = cheng(a, a);
            }

            f = cheng(f, temp); // 最后再用 {F(1,1) = 1 , F(1,2) = 1 } 乘上 (A^n - B)^(m - 1)
            cout << f.a[0][0] << endl;
        }
    }
    return 0;
}

---