给定一个整数数组 nums ,找到一个具有最大和的连续子数组(子数组最少包含一个元素),返回其最大和。
示例:
输入: [-2,1,-3,4,-1,2,1,-5,4],
输出: 6
解释: 连续子数组 [4,-1,2,1] 的和最大,为 6。
进阶:
如果你已经实现复杂度为 O(n) 的解法,尝试使用更为精妙的分治法求解。
利用分治算法求解最大子数组的问题
参考算法导论第4章
思路:如果我们要寻找子数组A[low,…high]的最大子数组。我们将数组分成两个同等规模的子数组。A[low,…,mid],A[mid+1,…high]。
现在,最大子数组A[i,j]必然处于以下三种情况之一:
- 完全位于子数组A[low,mid].
- 完全位于子数组A[mid+1.high].
- 跨越了中点,因此
l
o
w
≤
i
≤
m
i
d
≤
j
≤
h
i
g
h
low \leq i \leq mid \leq j \leq high
low≤i≤mid≤j≤high.
我们开始最大子数组跨越了终点的情况。下面给出跨越了中点的最大子数组算法。
def findMaxCrossingSubarray(A,low,mid,high):
left_sum = -float('inf')
sumnum = 0
for i in range(mid,low-1,-1):
sumnum = sumnum + A[i]
if sumnum > left_sum:
left_sum = sumnum
max_left = i # 求出A[low,..mid]最大子数组A[i,..mid]的索引i
right_sum = -float('inf')
sumnum = 0
for j in range(mid+1,high+1):
sumnum = sumnum + A[j]
if sumnum > right_sum:
right_sum = sumnum
max_right = j # 求出A[mid+1,..high]最大子数组A[mid+1,j]的索引j
return max_left,max_right,left_sum+right_sum
def findmaxsubarray(A,low,high):
if high < low:
return []
elif high == low:
return (low,high,A[low]) #如果只有一个元素就返回这个元素
else:
mid = int((low+high)/2)
(left_low,left_high,left_sum) = findmaxsubarray(A,low,mid) # 计算 A[low,mid]的最大子数组
(left_low,right_high,right_sum) = findmaxsubarray(A,mid+1,high) # 计算A[mid+1, high]的最大子数组
cross_low,cross_high,cross_sum = findmaxcrossingsubarray(A,low,mid,high) # 计算跨越中点的最大子数组
if left_sum >= right_sum and left_sum >= cross_sum: # 比较之后返回最大的
return (left_low,left_high,left_sum)
elif right_sum >= left_sum and right_sum >= cross_sum:
return (right_sum,right_high,right_sum)
else:
return cross_low,cross_high,cross_sum
#完整算法
class Solution:
def maxSubArray(self, nums):
"""
:type nums: List[int]
:rtype: int
"""
def findmaxsubarray(A,low,high):
if high < low:
return []
elif high == low:
return (low,high,A[low])
else:
mid = int((low+high)/2)
(left_low,left_high,left_sum) = findmaxsubarray(A,low,mid)
(left_low,right_high,right_sum) = findmaxsubarray(A,mid+1,high)
cross_low,cross_high,cross_sum = findmaxcrossingsubarray(A,low,mid,high)
if left_sum >= right_sum and left_sum >= cross_sum:
return (left_low,left_high,left_sum)
elif right_sum >= left_sum and right_sum >= cross_sum:
return (right_sum,right_high,right_sum)
else:
return cross_low,cross_high,cross_sum
def findmaxcrossingsubarray(A,low,mid,high):
left_sum = -float('inf')
sumnum = 0
for i in range(mid,low-1,-1):
sumnum = sumnum + A[i]
if sumnum > left_sum:
left_sum = sumnum
max_left = i
right_sum = -float('inf')
sumnum = 0
for j in range(mid+1,high+1):
sumnum = sumnum + A[j]
if sumnum > right_sum:
right_sum = sumnum
max_right = j
return max_left,max_right,left_sum+right_sum
return findmaxsubarray(nums,0,len(nums)-1)[-1]
利用动态规划的思想:
基于算法导论练习题4.1-5给出的思想:
若已知A[1…j]的最大子数组,基于如下性质将将拓展为A[1,j+1]的最大子数组:A[1…j+1]的最大子数组要么是A[1…j]的最大子数组,要么是某个子数组A[i,…j+1]。我们可以找到子问题即A[1]的最大子数组为A[1]。
我们用global[j]表示A[1…j]的最大子数组。
用local[j+1]表示最大子数组A[i,…j+1]。(即这个最大子数组必须包含A[j+1].)
我们可以写出子问题和递归式:
子问题:
l
o
c
a
l
[
1
]
=
A
[
1
]
;
g
l
o
b
a
l
[
1
]
=
A
[
1
]
local[1] = A[1]; global[1] = A[1]
local[1]=A[1];global[1]=A[1]
递归式:
l
o
c
a
l
[
j
+
1
]
=
m
a
x
(
l
o
c
a
l
[
j
]
+
A
[
j
+
1
]
,
A
[
j
+
1
]
)
g
l
o
b
a
l
[
i
+
1
]
=
m
a
x
(
g
l
o
b
a
l
[
i
]
,
l
o
c
a
l
[
j
+
1
]
)
local[j+1] = max(local[j] + A[j+1], A[j+1]) \\ global[i+1] = max(global[i], local[j+1])
local[j+1]=max(local[j]+A[j+1],A[j+1])global[i+1]=max(global[i],local[j+1])
class Solution:
def maxSubArray(self, nums):
"""
:type nums: List[int]
:rtype: int
"""
if len(nums) < 1:
return 0
globaln = nums[0]
local = nums[0]
for i in range(1,len(nums)):
local = max(nums[i],local+nums[i])
globaln = max(local,globaln)
return globaln