Numpy Arrays - Convert A 3D Array To A 2D Array

Having the following 3D array (9,9,9): np.arange(729).reshape((9,9,9)) [[[ 0 1 2 3 4 5 6 7 8] [ 9 10 11 12 13 14 15 16 17] [ 18 19 20 21 22 23 2

Solution 1:

You can firstly reshape the array to a 4d array, swap the second and third axises and then reshape it to 27 X 27:

a.reshape(3,3,9,9).transpose((0,2,1,3)).reshape(27,27)

#array([[  0,   1,   2,   3,   4,   5,   6,   7,   8,  81,  82,  83,  84,
#         85,  86,  87,  88,  89, 162, 163, 164, 165, 166, 167, 168, 169,
#        170],
#       [  9,  10,  11,  12,  13,  14,  15,  16,  17,  90,  91,  92,  93,
#         94,  95,  96,  97,  98, 171, 172, 173, 174, 175, 176, 177, 178,
#        179],
#  ... 
#       [558, 559, 560, 561, 562, 563, 564, 565, 566, 639, 640, 641, 642,
#        643, 644, 645, 646, 647, 720, 721, 722, 723, 724, 725, 726, 727,
#        728]])

Solution 2:

import numpy as np

x = np.arange(729).reshape(9, 9, 9)

y = x.transpose(1, 0, 2).reshape(27, 27)
y[y[:,2].argsort()]

Explanations

I used numpy.transpose to permute the stride and shape information for each axis.

>>> x.strides
(324L, 36L, 4L)
>>> x.transpose(1, 0, 2).strides
(36L, 324L, 4L)

more info in this answer.

Then I used numpy.reshape the 3D (9L, 9L, 9L) in 2D as expected

>>> x.reshape(27, 27)
(27L, 27L)

Of course, the combination of functions (like transpose and reshape) is very common in numpy. It allows you do to this matrix transformation in a one-liner:

x.transpose(1, 0, 2).reshape(27, 27)

EDIT

As @PaulPanzer point it out, the array was unsorted.

To sort array column by column, one can use:

y[y[:,2].argsort()]

But maybe it isn't the easiest answer anymore.

Solution 3:

If you are ok with moving data in a list you can use:

np.hstack([x for x in np.arange(729).reshape((9,9,9))])