Apply math.erf to an array

Question

I am using math.erf to find the error function of each element in an array.

# Import the erf function from the math library
from math import erf
# Import the numpy library to represent the input data
import numpy as np

# Create a dummy np.array
dummy_array = np.arange(20)

# Apply the erf function to the array to calculate the error function of each element
erf_array = erf(dummy_array)

I am getting an error as I cannot apply this whole function to an array. Is there a way to apply the error function to the whole array (vectorised approach) without looping through each element and applying it? (The loop will take a lot of time as the tables will be large)

The whole point of numpy that you don't have to use built-in math. What about [`scipy.special.erf`](https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.special.erf.html)? — Péter Leéh, Dec 01 '19 at 15:44
I return an unresolved reference in trying to call it and pycharm usually suggests functions that are available in the library — 78282219, Dec 01 '19 at 15:58
well, it's there. It's unusual that I don't get suggestions for it though like I usually would — 78282219, Dec 01 '19 at 15:59
@78282219 are you sure you have `scipy` installed on your system? Try `pip install scipy` — Code Different, Dec 01 '19 at 16:26
I found it but it doesn't show up on suggested, must be some reason — 78282219, Dec 01 '19 at 16:28

Stef · Accepted Answer · 2021-06-24T06:36:45.273

5

from scipy import special
import numpy as np

dummy_array = np.arange(20)
erf_array = special.erf(dummy_array)

It is essential that you import the special subpackage as from scipy import special. Importing just scipy as import scipy and then calling scipy.special.erf() won't work, as explained here and here.

edited Jun 24 '21 at 06:36

answered Dec 01 '19 at 19:28

Stef

28,728
2
24
52

I mean... you could just do `import scipy.special` then use `scipy.special.erf`. – Tadhg McDonald-Jensen Sep 19 '21 at 22:26
Is this actually faster than looping with math.erf()? – Anonymous Dec 05 '21 at 11:32
1

@Anonymous: yes, even for the given example with 20 elements I get 4 µs for `%timeit special.erf(dummy_array)` and 12 µs for `%timeit np.array([math.erf(element) for element in dummy_array])` (and 7 µs without array creation). For larger arrays the difference becomes bigger. – Stef Dec 05 '21 at 11:45

score 1 · Answer 2 · answered May 21 '21 at 12:39

1

you can use list comprehension to apply the function to each element at once erf_array = [erf(element) for element in dummy_array)]

answered May 21 '21 at 12:39

Ramzi

11
1

This will end up with a usual Python list, not a numpy array, which I presume OP wants implicitly. In this case the list should be fed to numpy’s `array` function, so that might be not as efficient due to reallocation. – arseniiv Jun 23 '21 at 18:04

Apply math.erf to an array

2 Answers2