Grouped AUC

Task

Implement grouped_auc(labels: list[int], scores: list[float], groups: list[int]) -> float. Compute AUC independently inside each valid group, then return the unweighted mean across groups that contain both classes.

Requirements

• Treat labels as binary values where 1 is positive and 0 is negative.
• Within each group, compare every positive score with every negative score.
• Award 1 for a correctly ordered pair, 0.5 for a tied pair, and 0 otherwise.
• Skip groups that do not contain at least one positive and one negative example.
• Give every valid group equal weight, regardless of its number of rows or pairs.
• Return 0.0 when there are no valid groups, and do not mutate the inputs.

Example

grouped_auc(
    [1, 0, 1, 0],
    [0.9, 0.2, 0.4, 0.4],
    [10, 10, 20, 20],
)
# 0.75