Numpy Arange generating values of inconsistent decimal points

Question

Can someone explain me what's happening here?

Why is there more decimal points for 0.3 and 0.7 values. I just want 1 decimal point values.

threshold_range = np.arange(0.1,1,0.1)
threshold_range.tolist()

[Output]: [0.1, 0.2, 0.30000000000000004, 0.4, 0.5, 0.6, 0.7000000000000001, 0.8, 0.9]

This is correct behavior. This occurs because python, like most programming languages, cannot EXACTLY represent some decimals due to floating point limitations. See also https://stackoverflow.com/questions/21895756/why-are-floating-point-numbers-inaccurate — tnknepp, Aug 01 '19 at 10:30
Possible duplicate of [Why are floating point numbers inaccurate?](https://stackoverflow.com/questions/21895756/why-are-floating-point-numbers-inaccurate) — tnknepp, Aug 01 '19 at 10:33
@tnknepp Thanks for that comment. That's the part I was curious about. — Saed, Aug 01 '19 at 11:25

score 5 · Accepted Answer · answered Aug 01 '19 at 10:27

5

Use np.round

Ex.

import numpy as np

threshold_range = np.arange(0.1,1,0.1)
print(threshold_range.tolist())
print(np.round(threshold_range, 2).tolist())

O/P:

[0.1, 0.2, 0.30000000000000004, 0.4, 0.5, 0.6, 0.7000000000000001, 0.8, 0.9]
[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]

answered Aug 01 '19 at 10:27

bharatk

Thanks for that. But I curious as to why it was acting up. Since you gave a solution to fix it first, I will mark it as the answer. As for the why, @tnknepp answered it in the comments. – Saed Aug 01 '19 at 11:22
@Saed [see this](https://stackoverflow.com/questions/12502122/python-numpy-floating-point-text-precision) Python/numpy floating-point text precision behaviour. – bharatk Aug 01 '19 at 11:46

score 3 · Answer 2 · answered Aug 01 '19 at 10:29

3

Solution: You can simply use round function:

threshold_range = np.arange(0.1,1,0.1).round(1)
threshold_range.tolist() # [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]

Reason of error: I think it has to do with floating point precision ;)

answered Aug 01 '19 at 10:29

Shariq farooq

Thanks for the answer. I had to mark the above one as the answer, as it that was the first response (just like yours) :) – Saed Aug 01 '19 at 11:24

2 Answers2