[Lintcode]Maximum Subarray III最大子数组 III

Given an array of integers and a number k, find k non-overlapping subarrays which have the largest sum.

The number in each subarray should be contiguous.

Return the largest sum.

Example

Given [-1,4,-2,3,-2,3], k=2, return 8

分析：与Best Time to Buy and Sell Stock IV类似，两个数组分别记录包含当前值的本地最优解和全局最优解。local[i][j]代表从0到nums中第i个数字处，分成j个数组后，和的最大值,且必须包括第i个数字，这样可以排除第i个数组包含在前一个数组中的情况。global[i][j]代表从0到nums第i个数字处，分成j个数组后和的最大值。

二维数组动态规划的状态转移方程为：

local[i][j] = Math.max(local[i - 1][j], global[i - 1][j - 1])  + nums[i - 1];

global[i][j] = Math.max(global[i - 1][j], local[i][j]);

public class Solution {
    /**
     * @param nums: A list of integers
     * @param k: An integer denote to find k non-overlapping subarrays
     * @return: An integer denote the sum of max k non-overlapping subarrays
     */
    public int maxSubArray(int[] nums, int k) {
        if(k > nums.length) return 0;
        
        int[][] local = new int[nums.length + 1][k + 1];
        int[][] global = new int[nums.length + 1][k + 1];
        //if k=0 local/global colomn 0 = 0
        //if nums.length=0 local/global row 0 = 0
        for(int i = 1; i <= nums.length; i++) {
            local[i][0] = Integer.MIN_VALUE;
            for(int j = 1; j <= k; j++) {
                if(j > i) {//矩阵中的值不能默认为0，否则影响结果
                    local[i][j] = Integer.MIN_VALUE;
                    global[i][j] = Integer.MIN_VALUE;
                    continue;
                }
                local[i][j] = Math.max(local[i - 1][j], global[i - 1][j - 1])  + nums[i - 1];
                if(i == j)
                    global[i][j] = local[i][j];
                else
                    global[i][j] = Math.max(global[i - 1][j], local[i][j]);
            }
        }
        
        return global[nums.length][k];
    }
}