Indexing a multi-dimensional array partially

Question

import numpy as np
arr = np.random.rand(50,3,3,3,16)
ids = (0,0,2,10)
b = arr[:, ids]  # don't work
b = arr[:, *ids]  # don't work
b = arr[:][ids]  # don't work
b = arr[:, tuple(ids)]  # don't work
b = arr[: + ids]  # don't work, obviously..
# b = arr[:,0,0,2,10].shape  # works (desired outcome)

I know there have been several questions about this, like Tuple as index of multidimensional array or Unpacking tuples/arrays/lists as indices for Numpy Arrays but none of them work for my case. Basically I want to index everything in the first axis, of specified "columns" in the rest of the axes (see the last line of my code). The desired output shape should be (50,) in this case.

But I want to index with a tuple/list of ids because I need to iterate through them, for example:

all_ids = ((0,0,0,2), (0,0,0,6), (1,1,0,2), (1,1,0,6),
           (2,2,0,2), (2,2,0,6), (2,2,2,2), (2,2,2,6))
c = 0
for id in all_ids:
    c += arr[:, id].sum()

akuiper · Accepted Answer · 2018-11-07 18:08:37Z

2

Add slice(None) to first dimension in ids and then subset:

arr[(slice(None),) + ids].shape
# (50,)

where:

(slice(None),) + ids
# (slice(None, None, None), 0, 0, 2, 10)

Notice slice(None, None, None) is equivalent to :, i.e. slice all. You can read docs on using slice object for indexing here.

edited Nov 7, 2018 at 18:08

answered Nov 7, 2018 at 18:02

akuiper

216k33 gold badges362 silver badges379 bronze badges

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

Collectives™ on Stack Overflow

Indexing a multi-dimensional array partially

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related