POJ 2923 Relocation(01背包变形, 状态压缩DP)

Q: 如何判断几件物品能否被 2 辆车一次拉走?

A: DP 问题. 先 dp 求解第一辆车能够装下的最大的重量, 然后计算剩下的重量之和是否小于第二辆车的 capacity, 若小于, 这 OK.

Description

Emma and Eric are moving to their new house they bought after returning from their honeymoon. Fortunately, they have a few friends helping them relocate. To move the furniture, they only have two compact cars, which complicates everything a bit. Since the furniture does not fit into the cars, Eric wants to put them on top of the cars. However, both cars only support a certain weight on their roof, so they will have to do several trips to transport everything. The schedule for the move is planed like this:

1. At their old place, they will put furniture on both cars.
2. Then, they will drive to their new place with the two cars and carry the furniture upstairs.
3. Finally, everybody will return to their old place and the process continues until everything is moved to the new place.

Note, that the group is always staying together so that they can have more fun and nobody feels lonely. Since the distance between the houses is quite large, Eric wants to make as few trips as possible.

Given the weights w_i of each individual piece of furniture and the capacities C₁ and C₂ of the two cars, how many trips to the new house does the party have to make to move all the furniture? If a car has capacity C, the sum of the weights of all the furniture it loads for one trip can be at most C.

Input

The first line contains the number of scenarios. Each scenario consists of one line containing three numbers n, C₁ and C₂. C₁ and C₂ are the capacities of the cars (1 ≤ C_i ≤ 100) and n is the number of pieces of furniture (1 ≤ n ≤ 10). The following line will contain n integers w₁, …, w_n, the weights of the furniture (1 ≤ w_i ≤ 100). It is guaranteed that each piece of furniture can be loaded by at least one of the two cars.

Output

The output for every scenario begins with a line containing “Scenario #i:”, where i is the number of the scenario starting at 1. Then print a single line with the number of trips to the new house they have to make to move all the furniture. Terminate each scenario with a blank line.

Sample Input

2
6 12 13
3 9 13 3 10 11
7 1 100
1 2 33 50 50 67 98

Sample Output

Scenario #1:
2

Scenario #2:
3

思路:

1. 题目给出, 一共 10 件家具, 所以可以使用状态压缩描述家具被装走的情况
2. 检查所有的状态(共 1<<n), 并判断其能够被 2 辆车一次带走, 可以的状态存储到 state 数组
3. 对 state 数组进行01 背包, 状态转移方程为 dp[s|state] = min(dp[s|state], dp[state]+1), 显然是个 push 的过程

总结:

1. 状态比较少时, 应该形成考虑状态压缩的思维定式, 就像 n<15 时可以暴力破解一样
2. 状态转移方程有 push 和 pull 两种, push 是指根据当前状态s来更新另一个状态 news; pull 是指通过另一个状态来更新当前状态
3. 在思路(3)的状态转移方程中可以看出, 当前状态是 state, 另一个状态是 s|state, 所以是 push
4. 处理状态压缩问题时, 读入的数组从 0 开始比较好, 因为 1<<I, I 总是从0 开始的
5. 代码第 48 行, 没判断 j &(state[i]) 导致 TLE 一次

代码:

#include <iostream>
using namespace std;

const int INF = 0X3F3F3F3F;
const int MAXN = 10 + 1<<11;
int N, C1, C2;
int a[15];
bool dp[MAXN];
int dp2[MAXN];
int state[MAXN];


bool check(const int &s) {
	memset(dp, 0, sizeof(dp));
	dp[0] = true;
	int sum = 0;
	for(int i = 0; i < N; i ++) {
		if(s & (1<<i)) { // 第 i 位为 1
			sum += a[i];
			for(int j = C1; j >= a[i]; j--) {	// 背包的容量是 C1, 01背包逆序遍历
				if(dp[j-a[i]])
					dp[j] = true;
			}
		}
	} 
	for(int i = 0; i <= sum; i++) { // 本来写成 C1
		if(dp[i] && sum-i <= C2) {
			return true;
		}
	}
	return false;
}

int solve_dp() {
	memset(state, 0, sizeof(state));
	int len = 0;
	int inf = 1<<N;
	for(int i = 0; i <= inf; i ++) {	// <=
		if(check(i))
			state[len++] = i;
	}

	// 第二次背包
	memset(dp2, 0x3f, sizeof(dp2));		// 初始化为 INF
	dp2[0] = 0;
	for(int i = 0; i < len; i ++) {
		for(int j = inf; j >= 0; j --) {
			if(!(j & (state[i]))) {
				dp2[j|state[i]] = min(dp2[j|state[i]], dp2[j]+1);
			}
		}
	}
	return dp2[(1<<N)-1];
}
int main() {
	freopen("E:\Copy\ACM\测试用例\in.txt", "r", stdin);
	int testcase;
	cin >> testcase;
	int tc = 0;
	while(testcase--) {
		tc++;
		scanf("%d%d%d", &N, &C1, &C2);
		for(int i = 0; i < N; i ++) {
			scanf("%d", &a[i]);
		}
		// solve
		printf("Scenario #%d:
%d

", tc, solve_dp());
	}
	return 0;
}

　　

update 2014年3月14日15:30:07

1. 总结 (5), 为什么不添加判断会超时? 假如不添加判断, 那么当 j&state[i] != 0 时也会执行状态转移方程. 考虑状态 state[k] 是 state[i] 的子集, 同时 j&state[k] == 0, 那么

j|state[i] == j|state[j] 同时 dp[j|state[i]] >= dp[j|state[k]], 所以对 j&state[i] != 0 的 case, 直接 continue 就好