如何转置 3D 矩阵？

Question

I have a 3D matrix x_test of size (100, 33, 66) and I want to change its dimensions to (100, 66, 33) .

What is the most efficient way to do this using python3.5? I look for something along those lines:

y = x_test.transpose()

Answer 1

You can pass the desired dimensions to the function np.transpose using in your case np.transpose(x_test, (0, 2, 1)) .

For example,

import numpy as np

x_test = np.arange(30).reshape(3, 2, 5)

print(x_test)
print(x_test.shape)

This will print

[[[ 0  1  2  3  4]
  [ 5  6  7  8  9]]

 [[10 11 12 13 14]
  [15 16 17 18 19]]

 [[20 21 22 23 24]
  [25 26 27 28 29]]]
(3, 2, 5)

Now, you can transpose the matrix with the command from above

y = np.transpose(x_test, (0, 2, 1))
print(y)
print(y.shape)

which will give

[[[ 0  5]
  [ 1  6]
  [ 2  7]
  [ 3  8]
  [ 4  9]]

 [[10 15]
  [11 16]
  [12 17]
  [13 18]
  [14 19]]

 [[20 25]
  [21 26]
  [22 27]
  [23 28]
  [24 29]]]
(3, 5, 2)

Answer 2

Apart from transpose (see @Cleb's answer) there are also swapaxes and moveaxis :

import numpy as np
mock = np.arange(30).reshape(2,3,5)

mock.swapaxes(1,2)
# array([[[ 0,  5, 10],
    [ 1,  6, 11],
    [ 2,  7, 12],
    [ 3,  8, 13],
    [ 4,  9, 14]],

   [[15, 20, 25],
    [16, 21, 26],
    [17, 22, 27],
    [18, 23, 28],
    [19, 24, 29]]])
np.moveaxis(mock,2,1)
# array([[[ 0,  5, 10],
    [ 1,  6, 11],
    [ 2,  7, 12],
    [ 3,  8, 13],
    [ 4,  9, 14]],

   [[15, 20, 25],
    [16, 21, 26],
    [17, 22, 27],
    [18, 23, 28],
    [19, 24, 29]]])

Answer 3

np.rot90 is another option. I confess I do not yet understand the axes = (a, b) notation and sort through all combinations from (0, 1) to (2, 1) to find what I want. Using x_test above, note its original shape (3, 2 ,5):

x2 = np.rot90(x_test, axes = (0, 1))

array([[[ 5,  6,  7,  8,  9],
    [15, 16, 17, 18, 19],
    [25, 26, 27, 28, 29]],

   [[ 0,  1,  2,  3,  4],
    [10, 11, 12, 13, 14],
    [20, 21, 22, 23, 24]]])

x2.shape
(2, 3, 5)