Replace -inf with zero value

Question

I have an array:

x =  numpy.array([-inf, -inf, 37.49668579])

Is there a way to change the -inf values to just 0?

score 105 · Accepted Answer · answered Jan 10 '14 at 16:56

105

There is:

from numpy import inf
x[x == -inf] = 0

answered Jan 10 '14 at 16:56

pv.

7

this is around 20% faster than np.isneginf based on my test – Warren Jun 20 '16 at 13:26
How are the brackets used here? Is this indexing with a conditional, or list comprehension? – chimeric Jul 25 '17 at 19:07
No list comprehension. It is numpy indexing using a boolean array. In this case: "Elements of x equal to -inf". – jabellcu Mar 27 '18 at 15:02
1

NameError: global name 'inf' is not defined – john k Nov 17 '18 at 23:02
@johnktejik did you run the import statement `from numpy import inf` ? – skjerns Jun 12 '20 at 09:06
I believe it's better to use `X[~np.isfinite(X)]` = 0 or `x[numpy.isneginf(x)] = 0` as in the second answer – Adiel Loinger Jul 06 '20 at 07:12

score 24 · Answer 2 · answered Jan 10 '14 at 17:00

24

x[numpy.isneginf(x)] = 0

answered Jan 10 '14 at 17:00

YXD

pr94 · Answer 3 · 2022-02-26T07:49:04.680

24

Perhaps even easier and more flexible to use numpy.nan_to_num:

numpy.nan_to_num(x, neginf=0) 

Out[1]: array([ 0.        ,  0.        , 37.49668579])

edited Feb 26 '22 at 07:49

answered Dec 18 '20 at 15:22

pr94

3 Answers3