Given a m x n
matrix mat
and an integer threshold
. Return the maximum side-length of a square with a sum less than or equal to threshold
or return 0 if there is no such square.
Example 1:
Input: mat = [[1,1,3,2,4,3,2],[1,1,3,2,4,3,2],[1,1,3,2,4,3,2]], threshold = 4
Output: 2
Explanation: The maximum side length of square with sum less than 4 is 2 as shown.
Example 2:
Input: mat = [[2,2,2,2,2],[2,2,2,2,2],[2,2,2,2,2],[2,2,2,2,2],[2,2,2,2,2]], threshold = 1
Output: 0
Example 3:
Input: mat = [[1,1,1,1],[1,0,0,0],[1,0,0,0],[1,0,0,0]], threshold = 6
Output: 3
Example 4:
Input: mat = [[18,70],[61,1],[25,85],[14,40],[11,96],[97,96],[63,45]], threshold = 40184
Output: 2
Constraints:
1 <= m, n <= 300
m == mat.length
n == mat[i].length
0 <= mat[i][j] <= 10000
0 <= threshold <= 10^5
-------------------------------------------------
前缀法很容易想到,但是到了二维,要先明确两点:
- 二维的前缀accu[x][y]=accu[x-1][y]+accu[x][y-1]-accu[x-1][y-1]+mat[x][y],简而言之就是行减一,列减一,再减去行列都减一重叠那部分,这种写法可以避免if else
- 前缀的差表示为arr[i]-arr[j],边长刚好是i-j,j可能是-1,在这种情况下直接让arr[j]为0,新定义一个if else来判断这种情况即可
所以,很容易想到几种解法:
解法A: 某一维上枚举,另外一维度上滑动窗口,这样的复杂度是min(rows,cols)^2*max(rows,cols)。
解法B: 二维循环枚举所有行列,利用1可以O(1)求出前缀和,利用2可以根据边长O(1)求出某个正方形的和,根据边长二分正方形的面积,让他<=阈值,相当于求最后一个<=,也就是upper_bound-1,复杂度为rows*cols*log(rows*cols),codes如下:
class Solution:
def maxSideLength(self, mat, threshold: int) -> int:
def accu(x,y):
return mat[x][y] if x>=0<=y else 0
rows,cols,edge = len(mat),len(mat[0]) if mat else 0,0
if (rows == 0 or cols == 0):
return 0
for i in range(rows):
for j in range(cols):
mat[i][j] += accu(i-1,j)+accu(i,j-1)-accu(i-1,j-1)
l,r = edge,min(i,j)+1
#upper_bound-1
while (l<=r):
mid = l + ((r-l)>>1)
fx = accu(i,j)-accu(i-mid,j)-accu(i,j-mid)+accu(i-mid,j-mid)
if (fx <= threshold):
l = mid+1
else:
r = mid-1
if (l-1>edge):
edge = l-1
return edge
s = Solution()
print(s.maxSideLength(mat = [[18,70],[61,1],[25,85],[14,40],[11,96],[97,96],[63,45]], threshold = 40184))
解法C:解法C是对B更好的改进,在新位置(i,j),尝试放比之前正方形边长edge大1的新正方形。为什么是比之前最好的edge大1呢?而不是大2呢?其实用反证法很容易证明。假设新正方形放得下edge+2,那么新正方形行-1、列-1的位置一定放得下edge+1的正方形,这个之前最好的正方形边长为edge矛盾。因此,只有edge+1有必要尝试,<edge+1的正方形没有必要试了,edge+2、edge+3。。这些更没有必要试了。。。这样时间复杂度成功降到O(rows*cols)
class Solution:
def maxSideLength(self, mat: List[List[int]], threshold: int) -> int:
def accu(x,y):
return mat[x][y] if x>=0<=y else 0
rows,cols = len(mat),len(mat[0]) if mat else 0
if (rows == 0 or cols == 0):
return 0
edge = 0
for i in range(rows):
for j in range(cols):
mat[i][j] += accu(i-1,j)+accu(i,j-1)-accu(i-1,j-1)
if (i>=edge<=j and accu(i,j)-accu(i-edge-1,j)-accu(i,j-edge-1)+accu(i-edge-1,j-edge-1)<=threshold):
edge += 1
return edge
-----------------------------------------------------------------------------------------------------
但是对于下面题目来说,就没法找到O(n^3)以下的解法了,老老实实利用前缀走起:
Given a matrix
, and a target
, return the number of non-empty submatrices that sum to target.
A submatrix x1, y1, x2, y2
is the set of all cells matrix[x][y]
with x1 <= x <= x2
and y1 <= y <= y2
.
Two submatrices (x1, y1, x2, y2)
and (x1', y1', x2', y2')
are different if they have some coordinate that is different: for example, if x1 != x1'
.
Example 1:
Input: matrix = [[0,1,0],[1,1,1],[0,1,0]], target = 0 Output: 4 Explanation: The four 1x1 submatrices that only contain 0.
Example 2:
Input: matrix = [[1,-1],[-1,1]], target = 0 Output: 5 Explanation: The two 1x2 submatrices, plus the two 2x1 submatrices, plus the 2x2 submatrix.
Note:
1 <= matrix.length <= 300
1 <= matrix[0].length <= 300
-1000 <= matrix[i] <= 1000
-10^8 <= target <= 10^8
from collections import defaultdict
class Solution:
def numSubmatrixSumTarget(self, matrix, target) -> int:
def A(x,y):
nonlocal a
return a[x][y] if x>=0<=y else 0
rows,cols = len(matrix),len(matrix[0]) if matrix else 0
b = matrix if (rows <= cols) else [[matrix[j][i] for j in range(rows)] for i in range(cols)]
rows,cols = len(b),len(b[0])
if (rows == 0 or cols == 0):
return 0
a = [[0 for j in range(cols)] for i in range(rows)]
for i in range(rows):
for j in range(cols):
a[i][j] = b[i][j]+A(i,j-1)+A(i-1,j)-A(i-1,j-1)
res = 0
for x1 in range(rows):
for x2 in range(x1,rows):
dic,px = defaultdict(int),0
dic[0] = 1 #bug2: forget this line
for y in range(cols):
px = A(x2,y)-A(x1-1,y) #bug1: px += A(x2,y)-A(x1-1,y
res += dic[px-target]
dic[px] += 1
return res