What's the most efficient way to replace some given indices of a NumPy array?

Question

I have three arrays, indices, values, and replace_values. I have to loop over indices, replacing each value in old_values[indices[i]] with new_values[i]. What's the fastest possible way to do this? It feels like there should be some way to speed it up using NumPy functions or advanced slicing instead of the normal for loop.

This code works, but is relatively slow:

import numpy as np

# Example indices and values
values = np.zeros([5, 5, 3]).astype(int)

indices = np.array([[0,0], [1,0], [1,3]])
replace_values = np.array([[140, 150, 160], [20, 30, 40], [100, 110, 120]])

print("The old values are:")
print(values)

for i in range(len(indices)):
    values[indices[i][0], indices[i][1]] = replace_values[i]

print("The new values are:")
print(values)

Sayandip Dutta · Accepted Answer · 2021-02-08 13:46:26Z

2

Use zip to separate x and y indices, then cast to tuple and assign:

>>> values[tuple(zip(*indices))] = replace_values
>>> values

array([[[140, 150, 160],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0]],

       [[ 20,  30,  40],
        [  0,   0,   0],
        [  0,   0,   0],
        [100, 110, 120],
        [  0,   0,   0]],

       [[  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0]],

       [[  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0]],

       [[  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0],
        [  0,   0,   0]]])

Where tuple(zip(*indices)) returns:

((0, 1, 1), (0, 0, 3))

As your indices is np.array itself, you can remove zip and use transpose, as pointed out by @MadPhysicist:

>>> values[tuple(*indices.T)]

edited Feb 8, 2021 at 13:46

answered Feb 8, 2021 at 12:46

Sayandip Dutta

15.9k4 gold badges27 silver badges57 bronze badges

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

1 Comment

Mad Physicist Over a year ago

tuple(zip(*indices)) would be much more effiently written as tuple(*indices.T). That way, you make a tuple from two arrays instead of converting each element into an int object.

Collectives™ on Stack Overflow

What's the most efficient way to replace some given indices of a NumPy 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