[FIXED] Avoid RuntimeWarning using where

Issue

I want to apply a function to a numpy array, which goes through infinity to arrive at the correct values:

def relu(x):
    odds = x / (1-x)
    lnex = np.log(np.exp(odds) + 1)
    return lnex / (lnex + 1)

x = np.linspace(0,1,10)
np.where(x==1,1,relu(x))

correctly computes

array([0.40938389, 0.43104202, 0.45833921, 0.49343414, 0.53940413,
       0.60030842, 0.68019731, 0.77923729, 0.88889303, 1.        ])

but also issues warnings:

3478817693.py:2: RuntimeWarning: divide by zero encountered in divide
  odds = x / (1-x)
3478817693.py:4: RuntimeWarning: invalid value encountered in divide
  return lnex / (lnex + 1)

How do I avoid the warnings?

Please note that performance is of critical importance here, so I would rather avoid creating intermediate arrays.

Solution

Another possible solution, based on np.divide, to avoid division by zero. This solution is inspired by @hpaulj's comment.

def relu(x):
    odds = np.divide(x, 1-x, out=np.zeros_like(x), where=x!=1)
    lnex = np.log(np.exp(odds) + 1)
    return lnex / (lnex + 1)

x = np.linspace(0,1,10)
np.where(x==1,1,relu(x))

Output:

array([0.40938389, 0.43104202, 0.45833921, 0.49343414, 0.53940413,
       0.60030842, 0.68019731, 0.77923729, 0.88889303, 1.        ])

Answered By - PaulS