Medium

Binary Cross-Entropy Loss

Medium

~15 min

code completion

For binary classification, the standard loss is binary cross-entropy (log loss):

L = - \frac{1}{m} i = 1 \sum m [y_{i} lo g (\overset{y}{^}_{i}) + (1 - y_{i}) lo g (1 - \overset{y}{^}_{i})]

where:

y_{i} \in {0, 1}

is the true label

\overset{y}{^}_{i} \in (0, 1)

is the predicted probability

Intuition:

When

y_{i} = 1

: penalizes low

\overset{y}{^}_{i}

via

- lo g (\overset{y}{^}_{i})

When

y_{i} = 0

: penalizes high

\overset{y}{^}_{i}

via

- lo g (1 - \overset{y}{^}_{i})

Your task:

Implement binary_cross_entropy(y_true, y_pred). Assume all y_pred values are strictly in (0, 1).

Example Tests

Mixed predictions, standard case

Input: {"y_pred":[0.9,0.1,0.8,0.2],"y_true":[1,0,1,0]}

Expected: 0.16425

Perfect confidence on correct label

Input: {"y_pred":[0.5],"y_true":[1]}

Expected: 0.69315

All correct with high confidence: low loss

Input: {"y_pred":[0.99,0.99,0.01,0.01],"y_true":[1,1,0,0]}

Expected: 0.01005

You can read the full problem statement above. Create a free account to run code in the browser, submit solutions, and track your progress.