numpy array with the second dimension empty

Question

Although I have seen some answers for this question, I was wondering if there's any better way to solve this problem.

For example,

N = 15
A = np.repeat(1/N, N)

if I do this the result will have shape (15, )

If I want the shape to be (15,1), I think I can do

A = np.repeat(1/N, N)[:,np.newaxis]

I also have the same kind of problems with

np.ones(N); np.zeros(N)

and I can probably make the second dimension as 1, by using "np.newaxis"

But my question is, are there any better ways to do this?

Using np.newaxis is a good tool for this task (though I use the shorter None). There's a minimal speed penalty. The idiom is clear (to experienced users). And it is useful in many other circumstances. — hpaulj
– hpaulj, Commented Aug 13, 2017 at 17:17
@hpaulj , please explain me why is my explanation is not optimal? — Ubdus Samad
– Ubdus Samad, Commented Aug 13, 2017 at 17:19
As written your reshape needs to know the length of A. There is a -1 shortcut, which often puzzles beginners. Timings for a small task like this are tricky (we are just making a view), but newaxis times faster. But for me it's mainly a matter of readability. The purpose of newaxis is clear. — hpaulj
– hpaulj, Commented Aug 13, 2017 at 17:36

Moses Koledoye · Accepted Answer · 2017-08-13 17:51:19Z

1

For np.ones and np.zeros you can pass a tuple specifying the shape of the output array:

np.ones((N, 1)); np.zeros((N, 1))

For np.repeat, you probably need to pass an array (or list) that already has two dimensions and then repeat along the desired axis:

>>> np.repeat([[1./N]], N, axis=0)
array([[ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667],
       [ 0.06666667]])

The syntax is, however, more difficult to read/understand, and promises no extra performances. You could stick to adding a new axis to the array like you've shown.

edited Aug 13, 2017 at 17:51

answered Aug 13, 2017 at 17:08

Moses Koledoye

78.8k8 gold badges139 silver badges141 bronze badges

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

2 Comments

hpaulj Over a year ago

Adding the extra dimensions to ones makes sense. Tweaking repeat like this is slower and harder to understand. np.full is faster.

Moses Koledoye Over a year ago

@hpaulj Thanks for noting. Didn't actually time it to see which works faster. Added some comments about this.

Collectives™ on Stack Overflow

numpy array with the second dimension empty

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related