2018-2019 ACM-ICPC, China Multi-Provincial Collegiate Programming Contest
We define $$$A$$$ is the permutation of n, and $$$B$$$ is a sequence of n
$$$Bi$$$ is defined as the number of subscripts that satisfy the condition $$$ k<i&&Ak>Ai $$$
Obviously,after a round of bubbling,$$$Bi$$$will minus one (if $$$Bi>0$$$), and $$$Ai$$$ will also move forward one space (i--)(although this is useless)
We consider to use sequence $$$B$$$ to solve the problem, we first prove that for every valid sequence $$$B$$$, there is exactly one permutation $$$A$$$ corresponding to it
We consider how to infer $$$A$$$ from $$$B$$$
It is very obvious that $$$An$$$ (the length of the permutation is n) can be directly infered by $$$Bn$$$, and $$$A n-1$$$ is also easily to know if $$$An$$$ is known,and so on. So as long as $$$B$$$ is legal, there is an $$$A$$$ permutation obtained, certificated
That is to say, for each bubbling, every $$$Bi$$$ greater than $$$0$$$ is decremented by one, and equal to $$$0$$$, it remains unchanged
So we only need to calculate the legal number of sequence of $$$B$$$
Consider what final states are legal:
All $$$Bi=0$$$ (ie $$$A$$$ is ordered)
There is a consecutive subsequence of $$$Bi=1$$$ in the permutation, and the rest are 0, such as 0 0 1 1 0 (1 4 2 3 5)
There is only one $$$Bi!=0$$$ in the permutation, and the others are 0, such as 0 0 2 0 0 (2 3 1 4 5)
I think it is easy to use Dynamic programming to solve this problem, set 3 states is ok
memset(dp,0,sizeof(dp)); int ans2=0; cin>>n>>m>>mod; dp[0][0]=1; for(int i=1;i<=n;i++){ if(i<=m+1){ (dp[i][0]+=dp[i-1][0]*(i))%=mod; } else{ (dp[i][0]+=dp[i-1][0]*(m+1))%=mod; (dp[i][1]+=dp[i-1][0]+dp[i-1][1])%=mod; (dp[i][2]+=dp[i-1][1]*(m+1)+dp[i-1][2]*(m+1)+dp[i-1][0]*max(0ll,i-1-(m+1)))%=mod; } } int ans=0; for(int i=0;i<3;i++)ans+=dp[n][i]; cout<<"Case #"<<ts<<": "<<ans%mod<<"\n";