https://www.lintcode.com/problem/knight-probability-in-chessboard/description
给定一个 N × N N\times N N×N的棋盘,一个国际象棋里的”马“从坐标 ( r , c ) (r,c) (r,c)出发,问其走 K K K步之后仍然在棋盘上的概率。要求中途不能出界。
法1:记忆化搜索。首先,设已经走过的步数是 s s s, A A A表示没走出界这个事件,那么由条件概率公式,得: P ( A ∣ s = k ) = 1 8 ∑ P ( A ∣ s = k + 1 ) P(A|s=k)=\frac{1}{8}\sum P(A|s=k+1) P(A∣s=k)=81∑P(A∣s=k+1)我们可以开一个记忆数组。设 f [ i ] [ j ] [ k ] f[i][j][k] f[i][j][k]代表已经走了 k k k步并且走到了 ( i , j ) (i,j) (i,j)的情况下,走完 K K K步没出界的概率。那么就有: f [ i ] [ j ] [ k ] = 1 8 ∑ f [ i ± 1 ] [ j ± 2 ] [ k + 1 ] + f [ i ± 2 ] [ j ± 1 ] [ k + 1 ] f[i][j][k]=\frac{1}{8}\sum f[i\pm 1][j\pm 2][k+1] + f[i\pm 2][j\pm 1][k+1] f[i][j][k]=81∑f[i±1][j±2][k+1]+f[i±2][j±1][k+1]边界条件,如果 k = K k=K k=K,那么 f f f的取值就取决于有没有出界,如果出界了,则取 0 0 0,否则取 1 1 1。代码如下:
import java.util.Arrays; public class Solution { /** * @param N: int * @param K: int * @param r: int * @param c: int * @return: the probability */ public double knightProbability(int N, int K, int r, int c) { // Write your code here. double[][][] dp = new double[N][N][K]; for (int i = 0; i < dp.length; i++) { for (int j = 0; j < dp[0].length; j++) { Arrays.fill(dp[i][j], -1); } } return dfs(r, c, N, 0, K, dp); } private double dfs(int x, int y, int N, int step, int K, double[][][] dp) { // 出界了,取0 if (outBound(x, y, N)) { return 0.0; } // 走完K步仍然没出界,条件概率等于1 if (step == K) { return 1.0; } // 有记忆则调取记忆 if (dp[x][y][step] >= 0) { return dp[x][y][step]; } // 开八个方向 int[] dx = {1, 2, 1, 2, -1, -2, -1, -2}, dy = {2, 1, -2, -1, 2, 1, -2, -1}; double prob = 0; for (int i = 0; i < 8; i++) { int nextX = x + dx[i], nextY = y + dy[i]; prob += 0.125 * dfs(nextX, nextY, N, step + 1, K, dp); } // 存一下记忆 dp[x][y][step] = prob; return prob; } private boolean outBound(int x, int y, int N) { return !(0 <= x && x < N && 0 <= y && y < N); } }时空复杂度 O ( N 2 K ) O(N^2K) O(N2K)。
法2:动态规划。设 f [ i ] [ j ] [ k ] f[i][j][k] f[i][j][k]是走 k k k步到达 ( i , j ) (i,j) (i,j)这个位置的概率。则 f [ i ] [ j ] [ k ] = 1 8 ∑ f [ i ± 1 ] [ j ± 2 ] [ k − 1 ] + f [ i ± 2 ] [ j ± 1 ] [ k − 1 ] f[i][j][k]=\frac{1}{8}\sum f[i\pm 1][j\pm 2][k-1] + f[i\pm 2][j\pm 1][k-1] f[i][j][k]=81∑f[i±1][j±2][k−1]+f[i±2][j±1][k−1]这个式子其实也是条件概率公式。那么最后的答案就是将 f [ 0 : N − 1 ] [ 0 : N − 1 ] [ k ] f[0:N-1][0:N-1][k] f[0:N−1][0:N−1][k]全加起来。代码如下:
public class Solution2 { /** * @param N: int * @param K: int * @param r: int * @param c: int * @return: the probability */ public double knightProbability(int N, int K, int r, int c) { // Write your code here. double[][][] dp = new double[N][N][K + 1]; dp[r][c][0] = 1.0; for (int k = 1; k <= K; k++) { for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { // 枚举八个方向 int[] dx = {1, 2, 1, 2, -1, -2, -1, -2}, dy = {2, 1, -2, -1, 2, 1, -2, -1}; for (int l = 0; l < 8; l++) { int prevX = i + dx[l], prevY = j + dy[l]; if (0 <= prevX && prevX < N && 0 <= prevY && prevY < N) { dp[i][j][k] += 0.125 * dp[prevX][prevY][k - 1]; } } } } } double res = 0.0; for (int i = 0; i < N; i++) { for (int j = 0; j < N; j++) { res += dp[i][j][K]; } } return res; } }时空复杂度 O ( N 2 K ) O(N^2K) O(N2K)。