HDU6198 number number number

We define a sequence F:
⋅⋅ F0=0,F1=1
⋅⋅ Fn=Fn1+Fn2(n2).

Give you an integer k , if a positive number n can be expressed by
n=Fa1+Fa2+...+Fak where 0a1a2ak , this positive number is mjfgood . Otherwise, this positive number is mjfbad .
Now, give you an integer k , you task is to find the minimal positive mjfbad number.
The answer may be too large. Please print the answer modulo 998244353.

Input
There are about 500 test cases, end up with EOF.
Each test case includes an integer kk which is described above. (1k109)

Output
For each case, output the minimal mjf−badmjf−bad number mod 998244353.

Sample Input
1

Sample Output
4


结论:求 fib0(2n+3) O(n) 的做法会超时,所以需要 O(logn) 的做法。
过程:暂时将「用一系列fibonacci数的和表示」这一操作称作「fib展开」,所用fibonacci数的数目最小的展开称作「最小展开」;题目要求的是「最小的·最小fib展开中fib数的数目大于k的数」。易知任意最小展开都不可能包含类似于 fib0(n)+fib0(n+1) 这样的东西,即所用的fib数一定是非相邻的;要求最小,所以符合要求的数的最小fib展开中任意两个fib数之间只会隔一个fib数。于是有两种可能:
1. fib0(0)+fib0(2)+fib0(4)+...
2. fib0(1)+fib0(3)+fib0(5)+...
但是第二种是不可能的。将第一种通过 fib0(n)=fib0(n1)+fib0(n2) 展开得到 fib0(0)+fib0(1)+fib0(2)+... ,即 fib0 前n项和;设为数列 Sfib0 ;考察可知这个数列的最小展开的fib数的数目为1 1 1 2 2 3 3 4 4 5 5… 猜测结果为 Sfib0(2k+1) ,再次观察数列猜测 Sfib0(2k+1)=fib0(2k+3)1
接下来就只剩下求 fib0 的问题了;500个testcase,k最大可以有 109 ,可以知道用 O(n) 的做法一定会TLE。网上面找到的做法都涉及矩阵快速幂;下面是一个来自这里的更加简单的做法:
fib0(2k)=fib0(k)[2fib0(k+1)fib0(k)
fib0(2k+1)=fib0(k+1)2+fib0(k)2
变换一下就是:
fib0(k)=fib0(k2)[2fib0(k2+1)fib0(k2)] (k is even)
fib0(k)=fib0(k+12)2+fib0(k12)2 (k is odd)
然后直接递归就好了。但是不能直接按公式做两次调用,那样会退化到会TLE的程度。
推理过程可以说是非常暴力了。也许好好证明之后结果也是这样吧。

#include<stdio.h>
#include<string.h>

typedef long long ll;

const int MOD = 998244353;
int K;

void quickfibx(int k, ll& a, ll& b){
    if(k==0){ a = 0, b = 1; return ; }
    if(k==1){ a = 1, b = 1; return ; }
    if(k==2){ a = 1, b = 2; return ; }
    quickfibx((k>>1),a,b);
    // now a = F(k>>1), b = F(k>>1 + 1);
    ll c = (a*((((2*b)%MOD)-a)%MOD))%MOD,
       d = (((a*a)%MOD)+((b*b)%MOD))%MOD;
    if(k%2==0){ a = c, b = d; }
    else { a = d, b = (c+d)%MOD; }
}
ll quickfib(int k){
    ll a, b;
    quickfibx(k,a,b);
    return a;
}

int main(){
    while(~scanf("%d",&K)){
        printf("%d\n",quickfib((K<<1)+3)-1);
    }
    return 0;
}
### HDU 1466 Problem Description and Solution The problem **HDU 1466** involves calculating the expected number of steps to reach a certain state under specific conditions. The key elements include: #### Problem Statement Given an interactive scenario where operations are performed on numbers modulo \(998244353\), one must determine the expected number of steps required to achieve a particular outcome. For this type of problem, dynamic programming (DP) is often employed as it allows breaking down complex problems into simpler subproblems that can be solved iteratively or recursively with memoization techniques[^1]. In more detail, consider the constraints provided by similar problems such as those found in references like HDU 6327 which deals with random sequences using DP within given bounds \((1 \leq T \leq 10, 4 \leq n \leq 100)\)[^2]. These types of constraints suggest iterative approaches over small ranges might work efficiently here too. Additionally, when dealing with large inputs up to \(2 \times 10^7\) as seen in reference materials related to counting algorithms [^4], efficient data structures and optimization strategies become crucial for performance reasons. However, directly applying these methods requires understanding how they fit specifically into solving the expectation value calculation involved in HDU 1466. For instance, if each step has multiple outcomes weighted differently based on probabilities, then summing products of probability times cost across all possible states until convergence gives us our answer. To implement this approach effectively: ```python MOD = 998244353 def solve_expectation(n): dp = [0] * (n + 1) # Base case initialization depending upon problem specifics for i in range(1, n + 1): total_prob = 0 # Calculate transition probabilities from previous states for j in transitions_from(i): # Placeholder function representing valid moves prob = calculate_probability(j) next_state_cost = get_next_state_cost(j) dp[i] += prob * (next_state_cost + dp[j]) % MOD total_prob += prob dp[i] %= MOD # Normalize current state's expectation due to accumulated probability mass if total_prob != 0: dp[i] *= pow(total_prob, MOD - 2, MOD) dp[i] %= MOD return dp[n] # Example usage would depend heavily on exact rules governing transitions between states. ``` This code snippet outlines a generic framework tailored towards computing expectations via dynamic programming while adhering strictly to modular arithmetic requirements specified by the contest question format.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值