使用numpy有效地测试矩阵行和列 [英] Efficiently test matrix rows and columns with numpy

问题描述

当第i行和第列都包含全0时，我试图同时删除第i行和第i列.例如，在这种情况下，我们可以看到第0行全为零，第0列全为零，因此删除了第0行和第0列.与第2列和第4列的行相同.第1行全为零，但第1列全为零，因此都不会删除.

I am trying to remove both the row i and column i when both the row i and column i contains all 0s. For example in this case we can see that row 0 is all zeros and column 0 is all zeros and thus row and column 0 is removed. Same with row column pair 2 and 4. Row 1 is all zeros but column 1 is not so neither are removed.

[0,0,0,0,0]
[0,1,0,1,0]
[0,0,0,0,0]
[0,0,0,0,0]
[0,0,0,0,0]

将成为

[1,1]
[0,0]

另一个例子:

[0,0,1,0,0,1]
[0,0,0,0,0,0]
[0,0,0,0,0,0]
[0,0,0,0,0,0]
[0,0,0,0,0,0]
[0,0,1,0,1,0]

将更改为:

[0,1,0,1]
[0,0,0,0]
[0,0,0,0]
[0,1,1,0]

这是我用来计算的代码:

This is the code that I am using to compute:

def remove(matrix):
    for i, x in reversed(list(enumerate(matrix))):
        if np.all(matrix == 0, axis=0)[i] and np.all(matrix == 0, axis=1)[i]:
            matrix = np.delete(matrix,i,axis=0)
            matrix = np.delete(matrix,i,axis=1)
    return matrix

经过测试，这条线是迄今为止最多的时间:

After testing this line is taking the most time by far:

if np.all(matrix == 0, axis=0)[i] and np.all(matrix == 0, axis=1)[i]:

是否有更合适的方法以这种方式测试行和列?我使用的矩阵是一个稀疏的二进制矩阵.我不使用任何稀疏矩阵类，只是ndarray.

Is there a more appropriate way to test a row and column in this way? The matrix that I am using is a sparse binary matrix. I am not using any sparse matrix classes just ndarray.

推荐答案

带遮罩的矢量化方法-

def remove_vectorized(a):
    mask = a==0
    m_row = ~mask.all(1)
    m_col = ~mask.all(0)
    comb_mask = m_row | m_col
    return a[comb_mask][:,comb_mask] #or a[np.ix_(comb_mask, comb_mask)]

示例运行

案例1:

In [485]: a
Out[485]: 
array([[0, 0, 0, 0, 0],
       [0, 1, 0, 1, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0]])

In [486]: remove_vectorized(a)
Out[486]: 
array([[1, 1],
       [0, 0]])

案例2:

In [489]: a
Out[489]: 
array([[0, 0, 1, 0, 0, 1],
       [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, 1, 0, 1, 0]])

In [490]: remove_vectorized(a)
Out[490]: 
array([[0, 1, 0, 1],
       [0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 1, 1, 0]])

这篇关于使用numpy有效地测试矩阵行和列的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！