Goods transportation

Description

There are $n$ cities located along the one-way road. Cities are numbered from $1$ to $n$ in the direction of the road.

The $i$-th city had produced $p_i$ units of goods. No more than $s_i$ units of goods can be sold in the $i$-th city.

For each pair of cities $i$ and $j$ such that $1 ≤ i < j ≤ n$ you can no more than once transport no more than $c$ units of goods from the city $i$ to the city $j$. Note that goods can only be transported from a city with a lesser index to the city with a larger index. You can transport goods between cities in any order.

Determine the maximum number of produced goods that can be sold in total in all the cities after a sequence of transportations.

Input

The first line of the input contains two integers $n$ and $c$ ($1 ≤ n ≤ 10 000$, $0 ≤ c ≤ 10^9$) — the number of cities and the maximum amount of goods for a single transportation.

The second line contains $n$ integers $p_i$ ($0 ≤ p_i ≤ 10^9$) — the number of units of goods that were produced in each city.

The third line of input contains n integers $s_i$ ($0 ≤ s_i ≤ 10^9$) — the number of units of goods that can be sold in each city.

Output

Print the maximum total number of produced goods that can be sold in all cities after a sequence of transportations.

Solution

Build a network of the following form: The network will consist of $n + 2$ vertices, the source located at the node with the number $n + 1$ and the sink at the node with the number $n + 2$. For each node $i$ such that $1 ≤ i ≤ n$ add the arc $(n + 1, i)$ with a capacity $p_i$ and arc $(i, n + 2)$ with capacity $s_i$. Also, for each pair of nodes $i$ and $j$ such that $1 ≤ i < j ≤ n$, add the arc $(i, j)$ with capacity m.

The value of the maximum flow in this network is the answer to the problem. Since the network contains $O(n^2)$ arcs, the usual flow search algorithms is not applicable here.

Instead, we will find a capacity of the minimal cut in this network. The cut in this case is a partition of the set of all nodes from $1$ to $n + 2$ into two parts. Let $A$ be the part that contains the source and — all other nodes. Then the capacity of the cut is equal to .

Capacity of the minimal cut can be found using the following dynamic programming: $cut(i, j)$ will be the minimum capacity of the cut, which was built on the first $i$ nodes, such that the set $A$ contains exactly $j$ nodes other than the source. It is easy to get the formula for this dynamic: $cut(i, j) = min(cut(i - 1, j - 1) + s_i, cut(i - 1, j) + p_i + j * c)$.

The answer to the problem can be obtained as $min_{j = 0, n}cut(n, j)$.

The resulting complexity of this solution is $O(n^2)$. To fit in memory limit you should store only two consecutive rows during the computation of the dynamic.

翻译一下就是考虑网络流：
1. $S$ to $i$ : $p_i$
2. $i$ to $T$ : $s_i$
3. $i$ to $j$ : $c$ $(i<j)$
最大流就是答案，但直接跑是 $O(n^2)$ 级别的边，肯定爆炸。考虑到这个网络的特殊性，同时考虑到最大流等于最小割。设计状态 $f[i][j]$ 表示对于前 $i$ 个点中有 $j$ 个点分到了 $S$ 集合的最小割边代价，转移为 $f[i][j]=min(f[i-1][j-1]+s_i,f[i-1][j]+p_i+j*c)$，转移时间复杂度 $O(n^2)$，空间上需要滚动一下。

Source

http://codeforces.com/contest/724/problem/E