## D - Zigzag Tree 解説 by evima

Let us choose the root arbitrarily so that the graph is rooted.

Assigning a permutation to the vertices can be rephrased into setting a total order on the vertices. The condition in question holds if and only if one of the following holds for any vertex \(v\) (the inequality is regarding the total order):

- \(v > w\) for any \(w\) adjacent to \(v\),
- \(v < w\) for any \(w\) adjacent to \(v\).

Let us say a \(v\) with the former property *big*, and a \(v\) with the latter property *small*.

### ◆ Tree DP

For a rooted subtree with \(n\) vertices rooted at \(v\), let us call the following value \(\mathrm{dp}_v[i]\).

- The number of ways to set a total order on the vertices in the rooted tree \(v\) such that the root \(v\) is small and the value assigned to the root is the \(i\)-th smallest value.

We can solve the problem by a tree DP that computes this from bottom to top. In this DP, we first need to merge the information from subtrees to compute the information for subforests in order.

For a forest with \(n\) vertices, we compute the following for each \(i\).

- The number of ways to set a total order on the vertices in the forest such that all roots are small and the maximum value assigned to the root is the \(i\)-th smallest value.

### ◆ Merging the DP table

We can see that computing the following is the key to the problem.

- We have a set \(A\) with \(n\) elements and a set \(B\) with \(m\) elements.
- \(\mathrm{dp}_A[i]\) is given, which is the number of total orders on \(A\) where the root is the \(i\)-th smallest and a certain requirement is satisfied. \(\mathrm{dp}_B[i]\) is given similarly.
- Compute \(\mathrm{dp}[i]\), the number of total orders on \(A\amalg B\) where the restrictions on \(A\) and \(B\) satisfy the requirement and the larger of the roots of \(A\) and \(B\) is the \(i\)-th smallest.

We will divide the computation into two, based on which of the roots is larger.

Let us fix the order \(a_1 < a_2 < \cdots < a_n\) on \(A\) and the order \(b_1 < b_2 < \cdots < b_m\) on \(B\). When the root of \(A\) is \(A_i\), the number of ways to extend the orders so that \(a_i\) is the \((A\amalg B)\)-th smallest in \(A\amalg B\) can be computed as the product of the following:

- setting the order between \(a_1,\ldots,a_{i-1}\), \(b_1,\ldots,b_j\): \(\binom{i-1+j}{i-1}\) ways
- setting the order between \(a_{i+1},\ldots,a_n\), \(b_{j+1},\ldots,b_m\): \(\binom{n-i+m-j}{n-i}\) ways

Additionally, \(a_i\) is larger than the root of \(B\) when the root of \(B\) is one of \(b_1,\ldots,b_j\). Thus, to \(\mathrm{dp}[i+j]\), we will add \(\binom{i-1+j}{i-1}\binom{n-i+m-j}{n-i}\mathrm{dp}_A[i] (\mathrm{dp}_B[1]+\cdots+\mathrm{dp}_B[j])\). After pre-calculation of prefix sums, this can be done in \(O(nm)\) time in total.

### ◆ Complexity Analysis

In a tree DP, if merging the information of subforests of sizes \(n\) and \(m\) takes \(O(nm)\) time, the information of the whole rooted tree with \(N\) vertices can be computed in \(O(N^2)\) time. The analysis of this complexity can be found, for example, on:

- ABC207 - F Editorial (https://atcoder.jp/contests/abc207/editorial/2155).

This analysis or tree DPs that involve it is often called “squared tree DP” in the Japanese kyo-pro community.

投稿日時:

最終更新: