NOTE

搬运文章，原创作者:http://joshuablog.herokuapp.com/
Just for study purpose, I don’t hold the copyright, if this is affecting anyone, please let me know.

概念

动态规划是通过拆分问题，定义问题状态和状态之间的关系，使得问题能够以递推（或者说分治）的方式去解决。
DP问题是Leetcode中的经典问题，也是面试中经常考到的类别之一，没有通用的模版，有些DP题思考的过程也比较繁琐.所以这篇总结可能会不断更新，以便达到更好的效果

适用场景

找Max， Min的问题
发现可能性的问题
输出所有解的个数问题

不适用场景

列出所有具体方案（起码是指数级别的复杂度，通常是递归，backtracking）
集合问题

考虑

状态
转移方程
初始化条件
返回结果
单序列问题
通常是数组，字符串的前N个为…

Warm Up

爬梯子

作为DP的入门题来说，思考过程还是很重要的。一次可以爬1级或者两级的台阶，问有多少种爬法。
符合输出所有解个数的问题。
因为只能爬一级或者两级所以到N级的话，你只能从n-1爬到n或者n-2爬到n；这样说来，如果dp[n]代表到n级台阶有多少种可能性的话，转移方程为dp[n] = dp[n-1]+dp[n-2]
所以代码很容易写出

def climb(n):
	if n == 1:
		return 1
	dp = [1 for _ in range(n+1)]
	dp[1] = 1
	dp[2] = 2
	for i in range(3,n+1):
		dp[i] = dp[i-1]+dp[i-2]
	return dp[-1]

当然这道题有优化条件，否则就没有必要花大篇幅写了。通过分析状态转移方程可以发现dp[i]只与dp[i-1],dp[i-2]有关，说明再之前的状态是不会影响到当前状态的，所以我们可以通过只保留两个状态来不断滚动从而求出最后的结果。

public class Solution {
    public int climbStairs(int n) {
        if (n == 1) {
            return 1;
        }
        int first = 1;
        int second = 2;
        for (int i = 3; i <= n; i++) {
            int third = first + second;
            first = second;
            second = third;
        }
        return second;
    }
}

进而我们能看出，如果当前状态只与前面的相关的话，我们都可以通过滚动数组，变量来简化空间复杂度–这种尤其适合不太复杂的动态规划问题，简单的二维DP

53. Maximum Subarray

找Max问题
很容易得出当前局部最大+当前值，和当前值的对比，而从决定是继续加还是从新来过

class Solution(object):
    def maxSubArray(self, nums):
        """
        :type nums: List[int]
        :rtype: int
        """
        if not nums:
            return 0
        if len(nums) == 1:
            return nums[0]
        dp = [float('-inf') for _ in range(len(nums))]

        dp[0] = nums[0]
        res = nums[0]
        for each in range(1,len(nums)):
            dp[each] = max(dp[each-1]+nums[each], nums[each])
            res = max(res, dp[each])
            #print dp
        return res

从上一段分析可以看出，dp状态只与上一个状态有关，从而可以简化成变量来储存dp[]

class Solution(object):
    def maxSubArray(self, nums):
        """
        :type nums: List[int]
        :rtype: int
        """
        if not nums:
            return 0
        if len(nums) == 1:
            return nums[0]
        prevMax = nums[0]
        res = nums[0]
        for each in range(1,len(nums)):
            prevMax = max(prevMax+nums[each], nums[each])
            res = max(res, prevMax)
            #print dp
        return res

300. Longest Increasing Subsequence

找出Max
这道题有点不一样的地方是最后的结果有可能是任意一个位置，所以不是简单的return dp[-1]而是max(dp)
dp[i] = 1 + max(dp[j]) j < i and A[i] > A[j]

class Solution(object):
    def lengthOfLIS(self, nums):
        """
        :type nums: List[int]
        :rtype: int
        """
        if not nums:
            return 0
        dp = [1] * (len(nums) + 1)
        for i in range(len(nums)):
            for j in range(i):
                if nums[j] < nums[i]:
                    dp[i] = max(dp[i], dp[j]+1)
        return max(dp)

139.Word Break

寻求解的存在性
和上一题有点像，dp[i] 为当前字符满足之前的字符在字典里

class Solution(object):
    def wordBreak(self, s, wordDict):
        """
        :type s: str
        :type wordDict: List[str]
        :rtype: bool
        """
        if not s:
            return False
        # dp保存dp【i】i之前的最少字符串
        dp = [False] * (len(s) + 1)
        dp[0] = True
        # 本身的循环，对字符串，在内层循环中需要使用i
        for i in range(1, len(s)+1):
            for j in range(i):
                if dp[j] and s[j:i] in wordDict:
                    dp[i] = True # 因为j～i是一个回文字符串
        return dp[-1]

198. House Robber

dp[i] = max(dp[i-2]+A[i], dp[i-1])
当然一共就3个状态，我们也可以通过类似爬梯子的方式，把空间复杂度降为O（1）

class Solution(object):
    def rob(self, nums):
        """
        :type nums: List[int]
        :rtype: int
        """
        if len(nums) == 0:
            return 0
        if len(nums) < 2:
            return nums[0] 
        dp = [0] * len(nums)
        dp[0] = nums[0]
        dp[1] = max(dp[0], nums[1])
        for i in range(2, len(nums)):
            dp[i] = max(dp[i-2] + nums[i], dp[i-1])
        return dp[-1]

303. Range Sum Query - Immutable

简单的累加求和做为DP，则转移方程为res(x,y) = dp[y] - dp[x-1]

class NumArray(object):

    def __init__(self, nums):
        """
        :type nums: List[int]
        """
        total = 0
        self.dp = []
        for i in nums:
            total += i
            self.dp.append(total)


    def sumRange(self, i, j):
        """
        :type i: int
        :type j: int
        :rtype: int
        """
        if i == 0:
            return self.dp[j]
        return self.dp[j] - self.dp[i-1]



# Your NumArray object will be instantiated and called as such:
# obj = NumArray(nums)
# param_1 = obj.sumRange(i,j)

Medium

368. Largest Divisible Subset

这道题需要一点思考，除了创建一个DP数组来记录到i位置最长的长度之外，我们还要知道其对应能整除的数字，所以还需要一个array来记录上一个数字的位置。dp的默认值为1， pre的初始值为None就是没有对应的数字；当且仅当dp需要更新的时候，更新其上一个能整除数字的index。最后找出max（dp）所对应的数字，根据其pre的index逐个找数字，输出。

class Solution(object):
    def largestDivisibleSubset(self, nums):
        """
        :type nums: List[int]
        :rtype: List[int]
        """
        if not nums:
            return []
        nums = sorted(nums)
        size = len(nums)
        dp = [1] * size
        pre = [None] * size
        for x in range(size):
            for y in range(x):
                if nums[x] % nums[y] == 0 and dp[y] + 1 > dp[x]:
                    dp[x] = dp[y] + 1
                    pre[x] = y
        idx = dp.index(max(dp))
        ans = []
        while idx is not None:
            ans += nums[idx],
            idx = pre[idx]
        return ans

338. Counting Bits

需要熟悉Bit运算和概念，要能发现countbit(n) = countbit(n/2) + n % 2这么一个方程，就是说一个数乘2意味着bit位左移一位

class Solution(object):
    def countBits(self, num):
        """
        :type num: int
        :rtype: List[int]
        """
        dp = [0 for _ in range(num+1)]
        for i in range(num+1):
            dp[i] = dp[i>>1] + i%2
        return dp

264. Ugly Number II

用三个dp存2，3，5出现作为乘子的个数

class Solution(object):
    def nthUglyNumber(self, n):
        """
        :type n: int
        :rtype: int
        """
        res = [1]
        i2, i3, i5 = 0,0,0
        while n:
            u2, u3, u5 = 2 *res[i2], 3*res[i3], 5*res[i5]
            temp = min(u2,u3,u5)
            if temp == u2:
                i2 += 1
            if temp == u3:
                i3 += 1
            if temp == u5:
                i5 += 1
            res.append(temp)
            n -= 1
        return res[-2]

673. Number of Longest Increasing Subsequence

相关题目，需要额外数组来记录已经出现最长的次数，也就是说如果前面有多个长度相等的连续子串的话，cnt要一直+1

class Solution(object):
    def findNumberOfLIS(self, nums):
        """
        :type nums: List[int]
        :rtype: int
        """
        if not nums:
            return 0
        dp = [1 for _ in range(len(nums))]
        dc = [1 for _ in range(len(nums))]
        for i in range(len(nums)):
            for j in range(i):

                if nums[i] > nums[j]:
                    if dp[i] < dp[j] + 1:
                        dp[i] = dp[j] + 1
                        dc[i] = dc[j]
                    elif dp[i] == dp[j] + 1:
                        dc[i] += dc[j] # we have multiple same length LIS, we need to add them
        res = 0

        for index, value in enumerate(dp):
            if value == max(dp):
                res += dc[index]
        return res

309. Best Time to Buy and Sell Stock with Cooldown

这道题一开始确实想不出来，感觉情况太多了；我承认有些DP题就是想不出来怎么做…..后来看到Discuss上有一个解法很不错，就是我们就考虑buy和sell的情况，buy的情况是最大利益只和前一个状态有关或者前两个状态的时候卖，然后这时候买。所以DP的定义就是buy[i] 在i的时候和在i之前买的最大值
buy[i] = Math.max(buy[i - 1], sell[i - 2] - prices[i]);
同理，sell的时候也和之前状态有关
sell[i] = Math.max(sell[i - 1], buy[i - 1] + prices[i]);

class Solution(object):
    def maxProfit(self, prices):
        """
        :type prices: List[int]
        :rtype: int
        """
        if not prices or len(prices) <= 1:
            return 0
        # buy[i] = max(buy[i-1], sell[i-2] - prices[i])
        # sell[i] = max(sell[i-1], buy[i-1] + prices[i])

        b0 = -prices[0]
        b1 = b0
        s0 = 0
        s1 = 0
        s2 = 0

        for i in range(1,len(prices)):
            b1 = max(b0, s0 - prices[i])
            s2 = max(s1, b0 + prices[i])

            b0 = b1
            s0 = s1
            s1 = s2
        return s2

矩阵DP

这种问题需要初始化DP数组，第0行和第0列，这样会方便之后的操作，通常这种问题是只能向右或者向下操作，否则则需要用BFS–求最短路径；DFS来解决

Warm Up

62. Unique Paths

符合求解的个数问题
dp[i][j] = dp[i-1][j] + dp[i][j-1]

class Solution(object):
    def uniquePaths(self, m, n):
        """
        :type m: int
        :type n: int
        :rtype: int
        """
        if not m or not n:
            return 0
        dp = [[1 for _ in range(n)] for _ in range(m)]
        for i in range(1,m):
            for j in range(1,n):
                dp[i][j] = dp[i-1][j] + dp[i][j-1]
        return dp[-1][-1]

当然我们知道每一个dp状态只与上方或者左方的状态相关，所以可以考虑通过某种方式来保存状态；一种方法是建立一个额外数组来存储列状态，dp变成一维存行状态；不过更进一步的话，每一个dp[i]状态就意味着当前从i行过来的状态总数，这里还是贴图吧，更好理解

class Solution(object):
    def uniquePaths(self, m, n):
        """
        :type m: int
        :type n: int
        :rtype: int
        """
        dp = [1 for _ in range(m)]
        for i in range(1,n):
            for j in range(1,m):
                dp[j] += dp[j-1]
        return dp[-1]

63.Unique Paths II

初始化首行首列的时候如果有障碍的话，就都变成0了

class Solution(object):
    def uniquePathsWithObstacles(self, obstacleGrid):
        """
        :type obstacleGrid: List[List[int]]
        :rtype: int
        """
        m = len(obstacleGrid)
        n = len(obstacleGrid[0])
        dp = [[0] * n] * m
        for i in range(m):
            for j in range(n):
                if obstacleGrid[i][j] == 1:
                    dp[i][j] = 0
                elif i == 0 and j == 0:
                    dp[i][j] = 1
                elif i == 0:
                    dp[i][j] = dp[i][j-1]
                elif j == 0:
                    dp[i][j] = dp[i-1][j]
                else:
                    dp[i][j] = dp[i-1][j] + dp[i][j-1]
        return dp[-1][-1]

64. Minimum Path Sum

class Solution(object):
    def minPathSum(self, grid):
        """
        :type grid: List[List[int]]
        :rtype: int
        """
        m = len(grid)
        n = len(grid[0])
        dp = [[0 for _ in range(n)] for _ in range(m)]
        dp[0][0] = grid[0][0]
        for i in range(1,m):
            dp[i][0] = dp[i-1][0] + grid[i][0]
        for j in range(1,n):
            dp[0][j] = dp[0][j-1] + grid[0][j]
        for i in range(1,m):
            for j in range(1,n):
                dp[i][j] = min(dp[i-1][j], dp[i][j-1]) + grid[i][j]
        #print dp
        return dp[-1][-1]

题目变成了正方形，逐层扩展的时候要记录，每一列计算之前要加上第0行第j列的值，所以沿用上一题的图，dp[i] = dp[i]–上方的值+ dp[i-1] – 左方的值

class Solution(object):
    def minPathSum(self, grid):
        """
        :type grid: List[List[int]]
        :rtype: int
        """
        m = len(grid)
        n = len(grid[0])
        dp = [0 for _ in range(m)]
        dp[0] = grid[0][0]
        for i in range(1,m):
            dp[i] = dp[i-1] + grid[i][0]

        for j in range(1,n):
            dp[0] = dp[0] + grid[0][j]
            for i in range(1,m):
                dp[i] = min(dp[i], dp[i-1]) + grid[i][j]
        #print dp
        return dp[-1]

256. Paint House

由于每次喷涂的房子颜色不能与之前的相同，所以转移方程为dp[i][2] += min(dp[i-1][0], dp[i-1][1]) dp为costs

class Solution(object):
    def minCost(self, costs):
        """
        :type costs: List[List[int]]
        :rtype: int
        """
        if not costs:
            return 0
        length = len(costs)
        for i in range(1, length):
            costs[i][0] += min(costs[i-1][1],costs[i-1][2])
            costs[i][1] += min(costs[i-1][0],costs[i-1][2])
            costs[i][2] += min(costs[i-1][0],costs[i-1][1])
        return min(costs[length-1][0], costs[length-1][1], costs[length-1][2])

276. Paint Fence

这道题其实跟上一道题很像，但是每一次涂得颜色可以和上一个一样（连续的颜色最多出现两次），所以当前状态与前两个有关，可以和前两个状态其中任何一个颜色一样 dp[2] = (k-1) * (dp[0] + dp[1]), dp = [k, k*k, 0]

class Solution(object):
    def numWays(self, n, k):
        """
        :type n: int
        :type k: int
        :rtype: int
        """
        if n == 0: return 0
        if n == 1: return k

        # for the first 2 posts
        dp = [k, k*k, 0]


        for i in range(2, n):
            dp[2] = (k-1) * (dp[0] + dp[1])
            dp[0] = dp[1]
            dp[1] = dp[2]
        return dp[1]

174. Dungeon Game

最主要的区别是，要从后往前找，由于生命最少为1，所以DP的条件也要相应变一下

class Solution(object):
    def calculateMinimumHP(self, dungeon):
        """
        :type dungeon: List[List[int]]
        :rtype: int
        """
        m = len(dungeon)
        n = len(dungeon[0])
        dp = [[0 for _ in range(n)] for _ in range(m)]

        for i in range(m-1,-1,-1):
            for j in range(n-1,-1,-1):
                if i == m-1 and j == n-1:
                    dp[i][j] = max(1, 1 - dungeon[i][j])
                elif i == m-1:
                    dp[i][j] = max(1, dp[i][j+1] - dungeon[i][j])
                elif j == n-1:
                    dp[i][j] = max(1, dp[i+1][j] - dungeon[i][j])
                else:
                    dp[i][j] = max(1, min(dp[i+1][j], dp[i][j+1]) - dungeon[i][j])
        return dp[0][0]

Medium

651. 4 Keys Keyboard

这道题想了半天，发现最后还是举例子最好；在6个操作内，只有A的操作是最大的，之后的话dp[i] = dp[j] * (i-j-1), 比如i==7， j==1的时候，最后是A * 7-1-1 = 5A

class Solution(object):
    def maxA(self, N):
        """
        :type N: int
        :rtype: int
        """
        if N <= 6:
            return N
        dp = [i for i in range(N+1)]
        for i in range(7,N+1):
            for j in range(1,i-2):
                dp[i] = max(dp[i], dp[j] * (i-j-1))
        return dp[-1]

304. Range Sum Query 2D - Immutable

这道题最关键的是处理corner case

class NumMatrix(object):

    def __init__(self, matrix):
        """
        :type matrix: List[List[int]]
        """
        if not matrix:
            return
        m = len(matrix)
        n = len(matrix[0])
        # deal with zero row and column
        self.dp = [[0 for _ in range(n+1)] for _ in range(m+1)]
        for i in range(1, m+1):
            for j in range(1, n+1):
                self.dp[i][j] = matrix[i-1][j-1] + self.dp[i-1][j] + self.dp[i][j-1] - self.dp[i-1][j-1]


    def sumRegion(self, row1, col1, row2, col2):
        """
        :type row1: int
        :type col1: int
        :type row2: int
        :type col2: int
        :rtype: int
        """
        return self.dp[row2+1][col2+1] - self.dp[row2+1][col1]- self.dp[row1][col2+1] + self.dp[row1][col1]



# Your NumMatrix object will be instantiated and called as such:
# obj = NumMatrix(matrix)
# param_1 = obj.sumRegion(row1,col1,row2,col2)

221. Maximal Square

可行的解法是很巧妙的：以这个square的最右下角的位置作为存储点f(i, j),当matrix(i, j)是1的时候，f(i, j) = min{f(i - 1, j - 1), f(i - 1, j), f(i, j -1)} + 1. 这是因为如果这是一个square，那么构成这个square的最基本条件就是跟它相邻的边的最小所在square.所以一个square的f值如下：

class Solution(object):
    def maximalSquare(self, matrix):
        """
        :type matrix: List[List[str]]
        :rtype: int
        """
        if not matrix:
            return 0
        res = 0
        m = len(matrix)
        n = len(matrix[0])
        dp = [[0 for _ in range(n+1)] for _ in range(m+1)]
        for i in range(1,m+1):
            for j in range(1, n+1):
                if matrix[i-1][j-1] == '1':

                    dp[i][j] = min(dp[i-1][j-1], dp[i-1][j], dp[i][j-1]) + 1
                    res = max(res, dp[i][j])

        return res * res

二维DP

72. Edit Distance

这道题是一道经典问题，分为之下三个操作，我觉得以下的解释是最好的
a) 插入一个字符：word1[0:i] -> word2[0:j-1]，然后在word1[0:i]后插入word2[j]
DP[i+1][j+1] = DP[i+1][j]+1

b) 删除一个字符：word1[0:i-1] -> word2[0:j]，然后删除word1[i]
DP[i+1][j+1] = DP[i][j+1]+1

c) 替换一个字符：word1[0:i-1] -> word2[0:j-1]
word1[i] != word2[j]时，word1[i] -> word2[j]：DP[i+1][j+1] = DP[i][j] + 1
word1[i] == word2[j]时：DP[i+1][j+1] = DP[i][j]

class Solution(object):
    def minDistance(self, word1, word2):
        """
        :type word1: str
        :type word2: str
        :rtype: int
        """
        l1 = len(word1)+1
        l2 = len(word2) + 1
        dp = [[0 for _ in range(l2)] for _ in range(l1)]

        for i in range(l1):
            dp[i][0] = i
        for j in range(l2):
            dp[0][j] = j
        for i in range(1,l1):
            for j in range(1,l2):
                if word1[i-1] == word2[j-1]:
                    dp[i][j] = dp[i-1][j-1]
                else:
                    dp[i][j] = min(dp[i][j-1],dp[i-1][j], dp[i-1][j-1]) + 1
        #print dp
        return dp[-1][-1]

DarkAlexWang's Blog

DP总结

概念

适用场景

考虑

单序列问题

Warm Up

爬梯子

53. Maximum Subarray

300. Longest Increasing Subsequence

139.Word Break

198. House Robber

303. Range Sum Query - Immutable

Medium

368. Largest Divisible Subset

338. Counting Bits

264. Ugly Number II

673. Number of Longest Increasing Subsequence

309. Best Time to Buy and Sell Stock with Cooldown

矩阵DP

Warm Up

62. Unique Paths

63.Unique Paths II

64. Minimum Path Sum

256. Paint House

276. Paint Fence

174. Dungeon Game

Medium

651. 4 Keys Keyboard

304. Range Sum Query 2D - Immutable

221. Maximal Square

二维DP

72. Edit Distance