使用 2d 数组索引 3d numpy 数组

Question

我想根据 numpy 3d 数组中的值创建一个 numpy 2d 数组，使用另一个 numpy 2d 数组来确定在轴 3 中使用哪个元素。

import numpy as np
#--------------------------------------------------------------------
arr_3d = np.arange(2*3*4).reshape(2,3,4)
print('arr_3d shape=', arr_3d.shape, '\n', arr_3d)
arr_2d = np.array(([3,2,0], [2,3,2]))
print('\n', 'arr_2d shape=', arr_2d.shape, '\n', arr_2d)
res_2d = arr_3d[:, :, 2]
print('\n','res_2d example using element 2 of each 3rd axis...\n', res_2d)
res_2d = arr_3d[:, :, 3]
print('\n','res_2d example using element 3 of each 3rd axis...\n', res_2d)

结果...

arr_3d shape= (2, 3, 4) 
 [[[ 0  1  2  3]
  [ 4  5  6  7]
  [ 8  9 10 11]]

 [[12 13 14 15]
  [16 17 18 19]
  [20 21 22 23]]]

 arr_2d shape= (2, 3) 
 [[3 2 0]
 [2 3 2]]

 res_2d example using element 2 of each 3rd axis...
 [[ 2  6 10]
 [14 18 22]]

 res_2d example using element 3 of each 3rd axis...
 [[ 3  7 11]
 [15 19 23]]

2 个示例结果显示了如果我使用轴 3 的第二个元素，然后使用第三个元素，我会得到什么。但我想从 arr_3d 获取由 arr_2d 指定的元素。所以...

- res_2d[0,0] would use the element 3 of arr_3d axis 3
- res_2d[0,1] would use the element 2 of arr_3d axis 3
- res_2d[0,2] would use the element 0 of arr_3d axis 3
etc

所以 res_2d 应该看起来像这样......

[[3 6 8]
[14 19 22]]

我尝试使用这一行来获取 arr_2d 条目，但它会生成一个 4 维数组，而我想要一个 2 维数组。

res_2d = arr_3d[:, :, arr_2d[:,:]]

Answer 1

花式索引和广播的结果的形状就是索引数组的形状。您需要为每个轴传递二维数组arr_3d

ax_0 = np.arange(arr_3d.shape[0])[:,None]
ax_1 = np.arange(arr_3d.shape[1])[None,:]

arr_3d[ax_0, ax_1, arr_2d]

Out[1127]:
array([[ 3,  6,  8],
       [14, 19, 22]])