Bernoulli NLL Gradient in Probability

Problem

For the Bernoulli negative log-likelihood L(p) = -y log(p) - (1 - y) log(1 - p), the derivative with respect to p is -y / p + (1 - y) / (1 - p).

Given p and y, return the derivative value.

Hint

Differentiate each log separately; one term vanishes depending on y.

case 1

in: [0.5,1]

expected: -2

case 2

in: [0.25,0]

expected: 1.3333333333333333

case 3

in: [0.8,1]

expected: -1.25

Starter codePython

def bernoulli_nll_gradient(p, y):
    # Your code here
    pass

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