Hinge Loss

Task

Implement hinge_loss(labels: list[int], scores: list[float], margin: float) -> float. Compute the mean binary hinge loss from -1/+1 labels and prediction scores.

Requirements

• labels and scores have the same shape, and every label is either -1 or +1.
• For each example, compute max(0, margin - label * score).
• Return the arithmetic mean of all per-example losses.
• Return 0.0 when the inputs are empty.
• An example exactly on the margin boundary contributes 0.0.
• Do not modify the inputs.

Example

labels = [1, -1, 1]
scores = [2.0, -0.5, 0.2]

hinge_loss(labels, scores, margin=1.0)
# 0.43333333333333335