Skip to main content
 首页 » 编程设计

python之填充由数组中的行、列索引偏移的对角线和反对角线之NumPy/Python

2024年11月24日11findumars

我有一个 n × n 矩阵。给定一个位置,我想用特定值替换该位置的对角线。

我尝试使用 np.fill_diagonal 但我没有得到正确的结果。

import numpy as np 
 
n = 8 
available = np.array([["O" for _ in range(n)] for _ in range(n)]) 
 
def update_available(column, row): 
    available[column][row] = "X" 
    np.fill_diagonal(available[column-row:],"X") 
    np.fill_diagonal(available[::-1,:][column-row-1:],"X") 
 

对于 update_available(4,1) 我想获得以下输出:

   0  1  2  3  4  5  6  7 
0  O  O  O  O  O  X  O  O 
1  O  O  O  O  X  O  O  O 
2  O  O  O  X  O  O  O  O 
3  X  O  X  O  O  O  O  O 
4  O  X  O  O  O  O  O  O 
5  X  O  X  O  O  O  O  O 
6  O  O  O  X  O  O  O  O 
7  O  O  O  O  X  O  O  O 

请您参考如下方法:

方法 #1:广播屏蔽

这是一个带有掩码的 -

def fill_cross(a, row, col, newval): 
    n = len(a) 
    r = np.arange(n) 
    m1 = r[::-1,None] == r + n-row-col-1 
    m2 = r[:,None] == r+row-col 
    a[m1|m2] = newval 
    return a 

方法 #2:NumPy-Eye-masking

将这些相同的偏移量放入 np.eye ,我们可以使它更紧凑,像这样-

def fill_cross_v2(a, row, col, newval): 
    n = len(a) 
    m1 = np.eye(n,k=-row+col,dtype=bool) 
    m2 = np.eye(n,k=n-col-row-1,dtype=bool)[:,::-1] 
    a[m1|m2] = newval 
    return a 

方法 #3:扁平化分配

我们还可以利用切片。这个想法是计算出对角线和反对角线的起始扁平化索引,并在扁平化的n+1(n =数组长度)步骤中对数组进行切片并分配新的值(value)观。这应该更有效,尤其是在大型阵列上。实现看起来像这样 -

def fill_cross_v3(a, row, col, newval): 
    if row+col>=n: 
        anti_diag_start = (row+col-n+1,n-1) 
    else: 
        anti_diag_start = (0,row+col) 
 
    if row>col: 
        diag_start = (row-col,0) 
    else: 
        diag_start = (0,col-row) 
 
    r,c = [np.ravel_multi_index(i,a.shape) for i in [diag_start,anti_diag_start]] 
    a.ravel()[r:r+(n-diag_start[0]-diag_start[1])*(n+1):n+1] = newval 
    a.ravel()[c:c*(n+1):n-1] = newval 
    return a 

sample 运行-

In [71]: a 
Out[71]:  
array([[0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0], 
       [0, 0, 0, 0, 0, 0, 0, 0]]) 
 
In [72]: fill_cross(a, row=4, col=1, newval=-1) 
Out[72]:  
array([[ 0,  0,  0,  0,  0, -1,  0,  0], 
       [ 0,  0,  0,  0, -1,  0,  0,  0], 
       [ 0,  0,  0, -1,  0,  0,  0,  0], 
       [-1,  0, -1,  0,  0,  0,  0,  0], 
       [ 0, -1,  0,  0,  0,  0,  0,  0], 
       [-1,  0, -1,  0,  0,  0,  0,  0], 
       [ 0,  0,  0, -1,  0,  0,  0,  0], 
       [ 0,  0,  0,  0, -1,  0,  0,  0]]) 

大型阵列上的计时 -

In [509]: a = np.zeros((1000,1000)) 
 
In [510]: %timeit fill_cross(a, row=200, col=700, newval=-1) 
     ...: %timeit fill_cross_v2(a, row=200, col=700, newval=-1) 
     ...: %timeit fill_cross_v3(a, row=200, col=700, newval=-1) 
1.64 ms ± 15.5 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) 
632 µs ± 787 ns per loop (mean ± std. dev. of 7 runs, 1000 loops each) 
438 µs ± 12.7 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) 
 
In [511]: a = np.zeros((5000,5000)) 
 
In [512]: %timeit fill_cross(a, row=2200, col=2700, newval=-1) 
     ...: %timeit fill_cross_v2(a, row=2200, col=2700, newval=-1) 
     ...: %timeit fill_cross_v3(a, row=2200, col=2700, newval=-1) 
66.6 ms ± 680 µs per loop (mean ± std. dev. of 7 runs, 10 loops each) 
39.1 ms ± 245 µs per loop (mean ± std. dev. of 7 runs, 10 loops each) 
17.3 ms ± 50.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)