Reputation: 613
In Matlab eps
has the following feature:
d = eps(x), where x has data type single or double, returns the positive distance from abs(x) to the next larger floating-point number of the same precision as x.
What is the equivalent way of computing this in Python or Numpy?
When searching for the answer, I found references to np.finfo(np.float64).eps
, which is only the equivalent of eps('double')
in Matlab.
Upvotes: 3
Views: 3664
Reputation: 1448
You might be searching for numpy spacing. Here an example:
import numpy as np
for i in [1e-2, 1, 1e5, 1e10]:
print(f'Spacing for {i:.4e} :\t {np.spacing(i):.4e}')
And here the output:
Spacing for 1.0000e-02 : 1.7347e-18
Spacing for 1.0000e+00 : 2.2204e-16
Spacing for 1.0000e+05 : 1.4552e-11
Spacing for 1.0000e+10 : 1.9073e-06
Upvotes: 2