numpy add variable number of dimensions to an array

Question

Is there a way to add a variable number of singleton dimensions to a numpy array? I want something like atleast_2d but for an arbitrary number of dimensions. expand_dims comes up a lot, but only adds a single dimension.

The only way I know to do this is to compute the shape first and apply it, i.e.

import numpy as np

def atleast_kd(array, k):
    array = np.asarray(array)
    new_shape = array.shape + (1,) * (k - array.ndim)
    return array.reshape(new_shape)

Looks good. That's just a variant on what expand_dims does: a.reshape(shape[:axis] + (1,) + shape[axis:]) — hpaulj
– hpaulj, Commented Feb 28, 2017 at 18:57
@hpaulj I should have looked at the source. This solution seems more elegant given that expand_dims doesn't do anything better. — Erik
– Erik, Commented Feb 28, 2017 at 19:07

zephyrus · Accepted Answer · 2020-05-25 04:24:49Z

1

A more elegant way to do this is np.broadcast_to. For example:

a = np.random.rand(2,2)
k = # number_extra dimensions
b = np.broadcast_to(a, (1,) * k + a.shape)

This will add the dimensions to the beginning of b. To get the exact same behavior as the given function, you can use np.moveaxis.

answered May 25, 2020 at 4:24

zephyrus

1,2661 gold badge12 silver badges31 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Erik Over a year ago

This works, but broadcast_to will do other things if you're not careful so it doesn't seem like the appropriate method, granted, so will reshape, so :/. It also doesn't fix the main issue of having to compute the "new shape".

Collectives™ on Stack Overflow

numpy add variable number of dimensions to an array

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related