Sampling Random Index from ndarray According to Array Elements

Question

I have an ndarray where the elements sum to 1. My problem is quite simple, I want to randomly sample indices of this array where the probability of getting each index is the value of the corresponding element. Now, I want to do this many times and in large arrays with many dimensions so I was hoping to use np.random.choice or something similar rather than a for-loop but I can't seem to figure it out.

For example, if I have the ndarray

import numpy as np
a = np.array([[0.0, 0.2], [0.6, 0.2]])

I want to never get index (0,0) get (0,1) and (1,1) at a rate of 20 % each and get (1,0) at a rate of 60 %.

Is there a good and fast way of doing this that you know of? Many thanks!

jlesuffleur · Accepted Answer · 2021-03-10 10:43:48Z

1

You can flatten the array (from 2D to 1D) and use np.random.choice. Then you need to project the 1D index back to 2D using np.unravel_index:

np.unravel_index(np.random.choice(a.size, p = a.flatten()), a.shape)

answered Mar 10, 2021 at 10:43

jlesuffleur

1,3031 gold badge9 silver badges20 bronze badges

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

2 Comments

Ivar Eriksson Over a year ago

This works. Thanks! I was going to comment and say that this is too slow for my purposes but np.random.choice has a size parameter and the time scales really well in this parameter. (For future reference)

Ivar Eriksson Over a year ago

This solution also works for arrays of an arbitrary dimension btw.

Collectives™ on Stack Overflow

Sampling Random Index from ndarray According to Array Elements

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