Softplus

Problem

Apply the softplus activation element-wise: softplus(x) = log(1 + exp(x)), computed as max(x, 0) + log1p(exp(-|x|)) so large magnitudes do not overflow.

The signature is softplus(x).

Hint

Softplus is a smooth ReLU; use log1p and the absolute value to stay stable.

Test cases

case 1

in: [[0]]

expected: [0.6931471806]

case 2

in: [[1,-1]]

expected: [1.3132616875,0.3132616875]

case 3

in: [[1000,-1000]]

expected: [1000,0]

case 4

in: [[2,3]]

expected: [2.126928011,3.0485873516]

Starter codePython

import math
def softplus(x):
    # Your code here
    pass

4 test cases · run your code against them in the editor