我有一个nxm个由非负整数组成的矩阵。例如:

2 3 4 7 1
1 5 2 6 2
4 3 4 2 1
2 1 2 4 1
3 1 3 4 1
2 1 4 3 2
6 9 1 6 4

“投下炸弹”会使目标细胞及其所有八个邻居的数量减少一个,直到最小值为零。

x x x 
x X x
x x x

什么样的算法可以确定将所有细胞减少到零所需的最少炸弹数量?

B选项(因为我不是一个细心的读者)

事实上,问题的第一个版本并不是我要寻找的答案。我没有仔细阅读整个任务,有额外的约束条件,让我们说:

那么简单的问题是,当行中的序列必须是非递增的:

8 7 6 6 5是可能的输入序列

7 8 5 5 2是不可能的,因为7 -> 8在一个序列中增长。

也许为“简单”的问题找到答案会有助于为更难的问题找到解决方案。

PS:我相信当我们有几个相同的情况需要最少的炸弹来清除上面的线时,我们会选择在“左侧”使用最多炸弹的一个。还有什么证据是正确的吗?


当前回答

这是另一个想法:

让我们先给黑板上的每个空格分配一个权重,计算在那里扔炸弹会减少多少数字。如果这个空间有一个非零数,它就得到一个点,如果它的相邻空间有一个非零数,它就得到一个额外的点。如果这是一个1000 * 1000的网格,我们为这100万个空间中的每一个都分配了权重。

然后根据权重对列表中的空格进行排序,并轰炸权重最高的空格。可以这么说,这是我们最大的收获。

在此之后,更新每个空间的重量是受炸弹的影响。这是你轰炸的空间,和它相邻的空间,以及它们相邻的空间。换句话说,任何空间的价值都可能因为爆炸而减少为零,或者相邻空间的价值减少为零。

然后,根据权重重新排序列表空间。由于轰炸只改变了一小部分空间的权重,因此不需要使用整个列表,只需在列表中移动这些空间。

轰炸新的最高权重空间,并重复上述步骤。

这保证了每次轰炸都能减少尽可能多的空格(基本上,它会击中尽可能少的已经为零的空格),所以这是最优的,除非它们的权重是相同的。所以你可能需要做一些回溯跟踪,当有一个平局的顶部重量。不过,只有最高重量的领带重要,其他领带不重要,所以希望没有太多的回溯。

Edit: Mysticial's counterexample below demonstrates that in fact this isn't guaranteed to be optimal, regardless of ties in weights. In some cases reducing the weight as much as possible in a given step actually leaves the remaining bombs too spread out to achieve as high a cummulative reduction after the second step as you could have with a slightly less greedy choice in the first step. I was somewhat mislead by the notion that the results are insensitive to the order of bombings. They are insensitive to the order in that you could take any series of bombings and replay them from the start in a different order and end up with the same resulting board. But it doesn't follow from that that you can consider each bombing independently. Or, at least, each bombing must be considered in a way that takes into account how well it sets up the board for subsequent bombings.

其他回答

到目前为止,一些答案给出了指数时间,一些涉及动态规划。我怀疑这些是否有必要。

我的解是O(mnS)其中m和n是板子的维度,S是所有整数的和。这个想法相当野蛮:找到每次可以杀死最多的位置,并在0处终止。

对于给定的棋盘,它给出28步棋,并且在每次落子后打印出棋盘。

完整的,不言自明的代码:

import java.util.Arrays;

public class BombMinDrops {

    private static final int[][] BOARD = {{2,3,4,7,1}, {1,5,2,6,2}, {4,3,4,2,1}, {2,1,2,4,1}, {3,1,3,4,1}, {2,1,4,3,2}, {6,9,1,6,4}};
    private static final int ROWS = BOARD.length;
    private static final int COLS = BOARD[0].length;
    private static int remaining = 0;
    private static int dropCount = 0;
    static {
        for (int i = 0; i < ROWS; i++) {
            for (int j = 0; j < COLS; j++) {
                remaining = remaining + BOARD[i][j];
            }
        }
    }

    private static class Point {
        int x, y;
        int kills;

        Point(int x, int y, int kills) {
            this.x = x;
            this.y = y;
            this.kills = kills;
        }

        @Override
        public String toString() {
            return dropCount + "th drop at [" + x + ", " + y + "] , killed " + kills;
        }
    }

    private static int countPossibleKills(int x, int y) {
        int count = 0;
        for (int row = x - 1; row <= x + 1; row++) {
            for (int col = y - 1; col <= y + 1; col++) {
                try {
                    if (BOARD[row][col] > 0) count++;
                } catch (ArrayIndexOutOfBoundsException ex) {/*ignore*/}
            }
        }

        return count;
    }

    private static void drop(Point here) {
        for (int row = here.x - 1; row <= here.x + 1; row++) {
            for (int col = here.y - 1; col <= here.y + 1; col++) {
                try {
                    if (BOARD[row][col] > 0) BOARD[row][col]--;
                } catch (ArrayIndexOutOfBoundsException ex) {/*ignore*/}
            }
        }

        dropCount++;
        remaining = remaining - here.kills;
        print(here);
    }

    public static void solve() {
        while (remaining > 0) {
            Point dropWithMaxKills = new Point(-1, -1, -1);
            for (int i = 0; i < ROWS; i++) {
                for (int j = 0; j < COLS; j++) {
                    int possibleKills = countPossibleKills(i, j);
                    if (possibleKills > dropWithMaxKills.kills) {
                        dropWithMaxKills = new Point(i, j, possibleKills);
                    }
                }
            }

            drop(dropWithMaxKills);
        }

        System.out.println("Total dropped: " + dropCount);
    }

    private static void print(Point drop) {
        System.out.println(drop.toString());
        for (int[] row : BOARD) {
            System.out.println(Arrays.toString(row));
        }

        System.out.println();
    }

    public static void main(String[] args) {
        solve();
    }

}

这里似乎有一个非二部匹配子结构。考虑下面的例子:

0010000
1000100
0000001
1000000
0000001
1000100
0010000

这种情况下的最佳解决方案的大小为5,因为这是9-cycle的边的最小顶点覆盖的大小。

这个例子,特别地,表明了一些人发布的线性规划松弛法是不精确的,不管用,还有其他一些不好的东西。我很确定我可以减少“用尽可能少的边覆盖我的平面立方图的顶点”来解决你的问题,这让我怀疑任何贪婪/爬坡的解决方案是否有效。

在最坏的情况下,我找不到在多项式时间内解出来的方法。可能有一个非常聪明的二进制搜索和dp解决方案,但我没有看到。

编辑:我看到这个比赛(http://deadline24.pl)是语言无关的;他们给你一堆输入文件,你给他们输出。所以你不需要在最坏情况下多项式时间内运行的东西。特别是,您可以查看输入!

There are a bunch of small cases in the input. Then there's a 10x1000 case, a 100x100 case, and a 1000x1000 case. The three large cases are all very well-behaved. Horizontally adjacent entries typically have the same value. On a relatively beefy machine, I'm able to solve all of the cases by brute-forcing using CPLEX in just a couple of minutes. I got lucky on the 1000x1000; the LP relaxation happens to have an integral optimal solution. My solutions agree with the .ans files provided in the test data bundle.

我敢打赌你可以用比我更直接的方式使用输入的结构,如果你看了它;看起来你只需要砍掉第一行,或者两行,或者三行直到你什么都不剩。(看起来,在1000x1000中,所有的行都是不增加的?我猜这就是你的“B部分”的来源吧?)

如果你想要绝对最优解来清理棋盘,你将不得不使用经典的回溯,但如果矩阵非常大,它将需要很长时间才能找到最佳解,如果你想要一个“可能的”最优解,你可以使用贪婪算法,如果你需要帮助写算法,我可以帮助你

现在想想,这是最好的办法。在那里制作另一个矩阵,存储通过投掷炸弹而移除的点,然后选择点数最多的单元格,并在那里投掷炸弹更新点数矩阵,然后继续。例子:

2 3 5 -> (2+(1*3)) (3+(1*5)) (5+(1*3))
1 3 2 -> (1+(1*4)) (3+(1*7)) (2+(1*4))
1 0 2 -> (1+(1*2)) (0+(1*5)) (2+(1*2))

对于每个相邻的高于0的单元格,单元格值+1

对于更新后的问题,简单的贪心算法可以得到最优结果。

向单元格A[1,1]投掷A[0,0]炸弹,然后向单元格A[2,1]投掷A[1,0]炸弹,并继续向下此过程。要清除左下角,向单元格A[n -2,1]投掷max(A[n -1,0], A[n -2,0], A[n -3,0])炸弹。这将完全清除前3列。

用同样的方法清除第3、4、5列,然后是第6、7、8列,等等。

不幸的是,这并不能帮助找到最初问题的解决方案。


“更大”的问题(没有“非增加”约束)可能被证明是np困难的。这是证明的草图。

假设我们有一个度为3的平面图形。我们来求这个图的最小顶点覆盖。根据维基百科的文章,这个问题对于3次以下的平面图形是np困难的。这可以通过平面3SAT的简化来证明。平面3SAT的硬度由3SAT降低而成。这两个证明都在Erik Demaine教授最近的“算法下界”讲座(第7和第9讲)中提出。

如果我们分割原始图的一些边(图中左边的图),每条边都有偶数个额外的节点,结果图(图中右边的图)应该对原始顶点具有完全相同的最小顶点覆盖。这样的转换允许将图顶点对齐到网格上的任意位置。

如果我们将图顶点只放置在偶数行和列上(这样就不会有两条边与一个顶点形成锐角),在有边的地方插入“1”,在其他网格位置插入“0”,我们可以使用原始问题的任何解决方案来找到最小顶点覆盖。

这是另一个想法:

让我们先给黑板上的每个空格分配一个权重,计算在那里扔炸弹会减少多少数字。如果这个空间有一个非零数,它就得到一个点,如果它的相邻空间有一个非零数,它就得到一个额外的点。如果这是一个1000 * 1000的网格,我们为这100万个空间中的每一个都分配了权重。

然后根据权重对列表中的空格进行排序,并轰炸权重最高的空格。可以这么说,这是我们最大的收获。

在此之后,更新每个空间的重量是受炸弹的影响。这是你轰炸的空间,和它相邻的空间,以及它们相邻的空间。换句话说,任何空间的价值都可能因为爆炸而减少为零,或者相邻空间的价值减少为零。

然后,根据权重重新排序列表空间。由于轰炸只改变了一小部分空间的权重,因此不需要使用整个列表,只需在列表中移动这些空间。

轰炸新的最高权重空间,并重复上述步骤。

这保证了每次轰炸都能减少尽可能多的空格(基本上,它会击中尽可能少的已经为零的空格),所以这是最优的,除非它们的权重是相同的。所以你可能需要做一些回溯跟踪,当有一个平局的顶部重量。不过,只有最高重量的领带重要,其他领带不重要,所以希望没有太多的回溯。

Edit: Mysticial's counterexample below demonstrates that in fact this isn't guaranteed to be optimal, regardless of ties in weights. In some cases reducing the weight as much as possible in a given step actually leaves the remaining bombs too spread out to achieve as high a cummulative reduction after the second step as you could have with a slightly less greedy choice in the first step. I was somewhat mislead by the notion that the results are insensitive to the order of bombings. They are insensitive to the order in that you could take any series of bombings and replay them from the start in a different order and end up with the same resulting board. But it doesn't follow from that that you can consider each bombing independently. Or, at least, each bombing must be considered in a way that takes into account how well it sets up the board for subsequent bombings.