Dynamically index a specific element in an n-dimensional numpy array

Question

I have an n-dimensional bool numpy array. I want to invert the bool value of a random item in the array. Inverting is easy, but I'm not sure how to best index a random element. I can generate a list of random positions along the n dimensions

indices = [np.random.randint(n) for n in array.shape]

However, how do I use that to index the corresponding element? array[indices] fetches elements at indices along the first dimension. array[*indices] does not work. I could do something like

a = array[indices[0]]
for index in indices[1:]:
    a = a[index]

but I'd like to avoid loops. Is there a more elegant solution?

[(*indices,)] will work fine

Mad Physicist
– Mad Physicist

2021-09-04 19:00:32 +00:00
Commented Sep 4, 2021 at 19:00 — Mad Physicist
– Mad Physicist, Commented Sep 4, 2021 at 19:00

Mark · Accepted Answer · 2021-09-04 19:14:51Z

2

I know this sound too simple, but use a tuple instead of a list for the index. There's some documentation about why this is in the Numpy docs, but (at least for me) the significance of tuples isn't immediately obvious (even though it was quite clear and there's a big warning box about this).

import numpy as np

np.random.seed(201)

A = np.random.choice(a=[False, True], size=(2,2))

# make it a Tuple
index = tuple(np.random.randint(n) for n in A.shape)

print(A)
# [[False True]
#  [False  True]]

print(index)
# (0, 1)

print(A[index])
# True

A[index] = ~A[index]
print(A)

# [[False  False]
#  [False  True]]

edited Sep 4, 2021 at 19:14

answered Sep 4, 2021 at 18:41

Mark

92.6k8 gold badges116 silver badges156 bronze badges

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

4 Comments

Mad Physicist Over a year ago

Why not just np.randint(a.shape)?

Mark Over a year ago

Yeah, @MadPhysicist that's a cleaner way to make the array. I suppose the main point of the answer is that would still need to be turned into a tuple to use it the way the OP wants and avoid triggering advanced indexing.

waterlemon Over a year ago

Thanks, that was it. I don't find this syntax to be very intuitive but I will sure remember it now.

hpaulj Over a year ago

It helps me to remember that a[3,4,2] is really a[(3,4,2)]. The index is a tuple, with one element per dimension. It's the comma that creates the tuple, the () are not always necessary.

Collectives™ on Stack Overflow

Dynamically index a specific element in an n-dimensional numpy array

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related