Shuffle a numpy array

Question

I have a 2-d numpy array that I would like to shuffle. Is the best way to reshape it to 1-d, shuffle and reshape again to 2-d or is it possible to shuffle without reshaping?

just using the random.shuffle doesn't yield expected results and numpy.random.shuffle shuffles only rows:

import random
import numpy as np
a=np.arange(9).reshape((3,3))
random.shuffle(a)
print a

[[0 1 2]
 [3 4 5]
 [3 4 5]]

a=np.arange(9).reshape((3,3))
np.random.shuffle(a)
print a

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

score 19 · Accepted Answer · answered Mar 15 '14 at 16:41

19

You can tell np.random.shuffle to act on the flattened version:

>>> a = np.arange(9).reshape((3,3))
>>> a
array([[0, 1, 2],
       [3, 4, 5],
       [6, 7, 8]])
>>> np.random.shuffle(a.flat)
>>> a
array([[3, 5, 8],
       [7, 6, 2],
       [1, 4, 0]])

answered Mar 15 '14 at 16:41

DSM

342,061
65
592
494

how to shuffle a matrix so that the correspondences between rows and columns are retained? For example, row *and* column 1 would be shuffled to row and column 3 in the shuffled matrix – develarist Sep 17 '20 at 13:15

score 8 · Answer 2 · answered Mar 15 '14 at 16:42

8

You could shuffle a.flat:

>>> np.random.shuffle(a.flat)
>>> a
array([[6, 1, 2],
       [3, 5, 0],
       [7, 8, 4]])

answered Mar 15 '14 at 16:42

NPE

486,780
108
951
1,012

Temak · Answer 3 · 2016-04-19T15:04:17.550

I think this is very important to note.
You can use random.shuffle(a) if a is 1-D numpy array. If it is N-D (where N > 2) than

random.shuffle(a)

will spoil your data and return some random thing. As you can see here:

import random
import numpy as np
a=np.arange(9).reshape((3,3))
random.shuffle(a)
print a

[[0 1 2]
 [3 4 5]
 [3 4 5]]

This is a known bug (or feature?) of numpy.

So, use only numpy.random.shuffle(a) for numpy arrays.

Shuffle a numpy array

3 Answers3

Linked