[Python] Why does axis=0 in np.mean() compute the mean across columns instead of rows?...

Why does axis=0 in np.mean() compute the mean across columns instead of rows?


Assume an example array like this:

import numpy as np

arr = np.array([[1,2,3], [1,2,3]])


>>> arr
array([[1, 2, 3],
[1, 2, 3]])
>>> arr.shape
(2, 3)


Since arr.shape[0] is 2, representing the number of rows, I expected np.mean(arr, axis=0) to compute the mean for each row, but instead, it computes the mean across columns:

>>> np.mean(arr, axis=0)
array([1., 2., 3.]) # Mean of each column
>>> np.mean(arr, axis=1)
array([2., 2.]) # Mean of each row


In the documentation, the "axis" parameter is described as:

"Axis or axes along which the means are computed."
​

I initially assumed that axis=0 would compute row-wise means returning the mean for each row because arr.shape[0] represents rows.

I am using python 3.10.0 and numpy version 2.2.3

Am I understanding the documentation wrong ? Could someone help clarify this convention? Why does axis=0 collapse rows instead of columns?

Continue reading...