Create Multidimensional Zeros Python

Question

I need to make a multidimensional array of zeros.

For two (D=2) or three (D=3) dimensions, this is easy and I'd use:

a = numpy.zeros(shape=(n,n)) 

or

a = numpy.zeros(shape=(n,n,n))

How for I for higher D, make the array of length n?

Solution 1

You can multiply a tuple (n,) by the number of dimensions you want. e.g.:

>>> import numpy as np
>>> N=2
>>> np.zeros((N,)*1)
array([ 0.,  0.])
>>> np.zeros((N,)*2)
array([[ 0.,  0.],
       [ 0.,  0.]])
>>> np.zeros((N,)*3)
array([[[ 0.,  0.],
        [ 0.,  0.]],

       [[ 0.,  0.],
        [ 0.,  0.]]])

Solution 2

>>> sh = (10, 10, 10, 10)
>>> z1 = zeros(10000).reshape(*sh)
>>> z1.shape
(10, 10, 10, 10)

EDIT: while above is not wrong, it's just excessive. @mgilson's answer is better.

Solution 3

In [4]: import numpy

In [5]: n = 2

In [6]: d = 4

In [7]: a = numpy.zeros(shape=[n]*d)

In [8]: a
Out[8]: 
array([[[[ 0.,  0.],
         [ 0.,  0.]],

        [[ 0.,  0.],
         [ 0.,  0.]]],


       [[[ 0.,  0.],
         [ 0.,  0.]],

        [[ 0.,  0.],
         [ 0.,  0.]]]])

Solution 4

you can make multidimensional array of zeros by using square brackets

array_4D = np.zeros([3,3,3,3])

View more solutions

Comments

[mgilson]

On second thought, I'm not really sure how this gets you anything that np.zeros doesn't already support. How is this better than np.zeros(sh)?

[mgilson]

This really isn't any different than my answer other than you're using a list rather than a tuple ...

[ev-br]

@mgilson: you're right. No need for extra brackets, as well: np.zeros(sh) works.

[mgilson]

Right, that's what I meant :)