Filling 3D numpy array with one variable dimension

Question

I would like to fill a 3D numpy array which has dimension (?, 100, 100). The first dimension will range between 500 and 1000 but is not known beforehand. I could use the append method, but this is slow.

Is there another posibility to create such a 3D numpy array where the first dimension is not known beforehand? The numpy array is filled using a for loop.

Second, let's assume I have the following two numpy arrays:

import numpy as np
arr1 = np.random.randint(0, 100, size=(30, 100, 100))
arr2 = np.random.randint(0, 100, size=(50, 100, 100))

How can I concatenate arr1 and arr2 along the first dimension so that the resulting array has shape (80, 100, 100)?

If you need to define the component arrays in a loop, collect them in a list, with list.append. Then do one concatenate (or maybe vstack or stack) at the end. — hpaulj
– hpaulj, Commented Jul 16, 2019 at 1:16
@hpaulj I have collected the arrays in a list, i.e. I have a list containing 2D arrays. When I use vstack or stack or concatenate over the list, then the arrays are concatenated either on the first or second dimension which I don't want. What I want is that when arrays in list have dimensions (a,b), I would like to get a resulting arrays with dimensions (n,a,b). Is this possible? — machinery
– machinery, Commented Jul 16, 2019 at 9:22

DeepBlue · Accepted Answer · 2019-07-15 23:02:29Z

2

About the first question, I always use np.empty and np.append methodes which are totaly fine.

arr = np.empty((0, 100, 100), int)
arr = np.append(arr,x,axis=0)

about the second question, append works well again:

arr3  = np.append(arr1, arr2, axis=0)

also 'concatenate' is usable:

arr3  = np.concatenate((arr1, arr2), axis=0)

answered Jul 15, 2019 at 23:02

DeepBlue

4464 silver badges9 bronze badges

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

2 Comments

hpaulj Over a year ago

np.append takes 2 arguments, tweaks them a bit, and then calls concatenate((arg1, arg2), axis=axis). It's ok when used once, but too many users try to use it in a loop. It's better to collect a list of arrays (list append is more efficient) and do one concatenate at the end. Plus defining that initial "empty" array is tricky for most beginners.

DeepBlue Over a year ago

About the storing data in a list and then make array from that I do agree that it is much faster.

Collectives™ on Stack Overflow

Filling 3D numpy array with one variable dimension

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