如何使用索引列表从 numpy arrays 列表中检索值？

Question

I have a list of numpy array indices which I created with argsort() :

i = 
[array([0, 1, 3, 2, 4], dtype=int64),
 array([1, 3, 0, 2, 4], dtype=int64),
 array([2, 4, 0, 1, 3], dtype=int64),
 array([3, 1, 0, 2, 4], dtype=int64),
 array([4, 2, 0, 3, 1], dtype=int64)]

This is the corresponding list of arrays with values:

v =
[array([0.        , 0.19648367, 0.24237755, 0.200832  , 0.28600039]),
 array([0.19648367, 0.        , 0.25492185, 0.15594099, 0.31378135]),
 array([0.24237755, 0.25492185, 0.        , 0.25685254, 0.2042604 ]),
 array([0.200832  , 0.15594099, 0.25685254, 0.        , 0.29995309]),
 array([0.28600039, 0.31378135, 0.2042604 , 0.29995309, 0.        ])] 

When I try to loop over the lists like this:

for line in i:
    v[line]

I get the error:

TypeError: only integer scalar arrays can be converted to a scalar index

But when I try to access them individually like this:

v[0][i[0]]

It works and outputs the values in v[0] in correct order like this:

array([0.        , 0.19648367, 0.200832  , 0.24237755, 0.28600039])

I want the arrays in v ordered from the smallest value to biggest. What am I doing wrong?

Answer 1

Loop through each line of i, and loop through each line of v at the same time using enumerate:

import numpy as np 

i = np.array([[0, 1, 3, 2, 4], [1, 3, 0, 2, 4], [2, 4, 0, 1, 3], [3, 1, 0, 2, 4], [4, 2, 0, 3, 1]])

v = np.array([[0.        , 0.19648367, 0.24237755, 0.200832  , 0.28600039], 
[0.19648367, 0.        , 0.25492185, 0.15594099, 0.31378135],
[0.24237755, 0.25492185, 0.        , 0.25685254, 0.2042604 ],
[0.200832  , 0.15594099, 0.25685254, 0.        , 0.29995309],
[0.28600039, 0.31378135, 0.2042604 , 0.29995309, 0.        ]] )

# you can rearrange each line of v by using indices in each row of i
for index, line in enumerate(i):
    print(v[index][line])

Output:

[0.         0.19648367 0.200832   0.24237755 0.28600039]
[0.         0.15594099 0.19648367 0.25492185 0.31378135]
[0.         0.2042604  0.24237755 0.25492185 0.25685254]
[0.         0.15594099 0.200832   0.25685254 0.29995309]
[0.         0.2042604  0.28600039 0.29995309 0.31378135]

Answer 2

This is all easier (and faster) if you don't use a python list of Numpy arrays, but instead use a multi-dimensional numpy array. Then you have all the great tool from numpy at you disposal and can avoid slow loops. For example for you can use np.take_along_axis :

from numpy import array 

i = np.array([
    [0, 1, 3, 2, 4],
    [1, 3, 0, 2, 4],
    [2, 4, 0, 1, 3],
    [3, 1, 0, 2, 4],
    [4, 2, 0, 3, 1]])


v = array([
    [0., 0.19648367, 0.24237755, 0.200832  , 0.28600039],
    [0.19648367, 0.        , 0.25492185, 0.15594099, 0.31378135],
    [0.24237755, 0.25492185, 0.        , 0.25685254, 0.2042604 ],
    [0.200832  , 0.15594099, 0.25685254, 0.        , 0.29995309],
    [0.28600039, 0.31378135, 0.2042604 , 0.29995309, 0.        ]] 
)


np.take_along_axis(v,i, 1)

result:

array([[0.        , 0.19648367, 0.200832  , 0.24237755, 0.28600039],
       [0.        , 0.15594099, 0.19648367, 0.25492185, 0.31378135],
       [0.        , 0.2042604 , 0.24237755, 0.25492185, 0.25685254],
       [0.        , 0.15594099, 0.200832  , 0.25685254, 0.29995309],
       [0.        , 0.2042604 , 0.28600039, 0.29995309, 0.31378135]])