How to set very small numbers to zero in numpy arrays?

Question

What is the best way to set numbers smaller than a specified tolerance to zero in a real numpy array? In particular, my array has 3 dimensions.

import numpy as np

N = 25
D = 50
tolerance = 10**-2

X = np.random.normal(0, 1, (N, D, 4))

I would like to set to 0 all values of X smaller than the tolerance set.

Does this answer your question? [Replacing Numpy elements if condition is met](https://stackoverflow.com/questions/19766757/replacing-numpy-elements-if-condition-is-met) — Zuku, Mar 07 '20 at 14:06
Small in amplitude, right? Because e.g. -100 is smaller than 10**-2, but you don't want to set it to 0. — Andreas K., Mar 07 '20 at 15:17

score 4 · Answer 1 · answered Mar 07 '20 at 15:29

4

I think a better way would be to use np.isclose

X[np.isclose(X, 0, atol=tolerance)] = 0

answered Mar 07 '20 at 15:29

FBruzzesi

score 3 · Accepted Answer · edited Mar 07 '20 at 15:22

3

This should do the trick:

X[X < tolerance] = 0

edited Mar 07 '20 at 15:22

Andreas K.

answered Mar 07 '20 at 14:05

cn1adil

2

-10 would meet that condition – Mad Physicist Mar 07 '20 at 14:06
It shouldn't? why tho? – cn1adil Mar 07 '20 at 14:09
A normal distribution is symmetrical about zero – Mad Physicist Mar 07 '20 at 14:10
But how is my answer incorrect against the OP's question? – cn1adil Mar 07 '20 at 14:18
6

Just use `X[abs(X) < tolerance] = 0`. – Andreas K. Mar 07 '20 at 15:15
whether the abs() is needed or not really depends on what you read, the question's header or text... ;-) – Nyps Oct 22 '20 at 11:25

2 Answers2