Cross-Entropy Loss (One-Hot)

Problem

Compute the mean cross-entropy loss in nats for one-hot labelled samples. probs is a list of probability rows and labels holds the index of the correct class for each row. Each selected probability is floored at 1e-12, and the function returns 0.0 for empty input.

Hint

Average -ln(p[label]) over all samples.

Test cases

case 1

in: [[[0.8,0.2],[0.1,0.9]],[0,1]]

expected: 0.164252033486018

case 2

in: [[[0.5,0.5],[0.25,0.75]],[1,0]]

expected: 1.0397207708399179

case 3

in: [[[1,0]],[0]]

expected: 0

case 4

in: [[],[]]

expected: 0

Starter codePython

import math
def cross_entropy_one_hot(probs, labels):
    # Your code here
    pass

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