[FIXED] Get element-wise extreme values of multiple arrays

Issue

Given multiple numpy arrays of same size, I want to create an array of equal size, which contains the extreme values of the arrays.

I have a working solution, but want to do this in a more pythonic way without the for loops if possible.

MWE

import numpy as np

a = np.array([[-1, 2], [3, 4]])
b = np.array([[-3, 1], [5, -2]])
c = np.array([[2, -1], [1, -5]])

m, n = a.shape

out = np.zeros((m, n))
for ii in range(m):
    for jj in range(n):
        tmp = [a[ii, jj], b[ii, jj], c[ii, jj]]
        out[ii, jj] = tmp[np.argmax(np.abs(tmp))]

print(out)
# [[-3, 2], [5, -5]]

I have been thinking about a combination of numpy.maximum.reduce and numpy.minimum.reduce, but were not able to reproduce the above result.

Solution

You can use the following:

w = np.dstack([a,b,c])
y = np.abs(w)
x = np.argmax(y, axis=-1)

z = np.take_along_axis(w, np.expand_dims(x, axis=-1), axis=-1)
np.reshape(z, (2,2))

array([[-3, 2],
       [5, -5]])

Answered By - C.Nivs