这是一个令人讨厌的方式,在一行中做到这一点!
A.reshape((-1, 3, 3)).swapaxes(-1, 1).reshape(A.shape)
一步步.重塑为(3,3,3)
>>> A.reshape((-1, 3, 3))
array([[[ 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]]])
然后在每个子阵列上执行类似转置的操作swapaxes
>>> A.reshape((-1, 3, 3)).swapaxes(-1, 1)
array([[[ 1, 4, 7],
[ 2, 5, 8],
[ 3, 6, 9]],
[[10, 13, 16],
[11, 14, 17],
[12, 15, 18]],
[[19, 22, 25],
[20, 23, 26],
[21, 24, 27]]])
最后重塑为(9,3).
>>> A.reshape((-1, 3, 3)).swapaxes(-1, 1).reshape(A.shape)
array([[ 1, 4, 7],
[ 2, 5, 8],
[ 3, 6, 9],
[10, 13, 16],
[11, 14, 17],
[12, 15, 18],
[19, 22, 25],
[20, 23, 26],
[21, 24, 27]])
>>>
我认为,对于任何方法,必须复制数据,因为没有2d步幅/形状可以生成以下结果:
array([ 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])
(是吗?)在我的版本中,我认为数据会在最终的重塑步骤中被复制