整形重塑/尺寸改变

Question

I want to change an array such as

array([[  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.]])

To

array([[[7.,   3.],
        [7.,   3.],
        [7.,   3.],
        [7.,   3.]],

       [[14.,   1.],
        [14.,   1.],
        [14.,   1.],
        [14.,   1.]],

       [[9.,   17.],
        [9.,    17.],
        [9.,   17.],
        [9.,   17.]]])

I thought I would manage with reshape, but none of the order statements work in this case. Currently, I am doing it this way

np.vstack([mat[:, i-2:i] for i in range(2, mat.shape[1]+1, 2)]).reshape(3,-1,2)

And I was wondering if there is a better way to do it

Answer 1

reshape and then swapaxes :

import numpy as np

a = np.array(
      [[  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.],
       [  7.,   3.,  14.,   1.,   9.,  17.]])

a.reshape((a.shape[0], -1, 2)).swapaxes(0, 1)