Metrics

Base

fairlearn_fhe.metrics.selection_rate(y_true, y_pred, *, pos_label=1, sample_weight=None)

fairlearn_fhe.metrics.mean_prediction(y_true, y_pred, sample_weight=None)

fairlearn_fhe.metrics.true_positive_rate(y_true, y_pred, sample_weight=None, pos_label=None)

fairlearn_fhe.metrics.true_negative_rate(y_true, y_pred, sample_weight=None, pos_label=None)

fairlearn_fhe.metrics.false_positive_rate(y_true, y_pred, sample_weight=None, pos_label=None)

fairlearn_fhe.metrics.false_negative_rate(y_true, y_pred, sample_weight=None, pos_label=None)

fairlearn_fhe.metrics.count(y_true, y_pred)

Group size — independent of whether y_pred is encrypted.

Fairness

fairlearn_fhe.metrics.demographic_parity_difference(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None)

fairlearn_fhe.metrics.demographic_parity_ratio(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None)

fairlearn_fhe.metrics.equalized_odds_difference(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None, agg='worst_case')

fairlearn_fhe.metrics.equalized_odds_ratio(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None, agg='worst_case')

fairlearn_fhe.metrics.equal_opportunity_difference(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None)

fairlearn_fhe.metrics.equal_opportunity_ratio(y_true, y_pred, *, sensitive_features, method='between_groups', sample_weight=None)

Derived

fairlearn_fhe.metrics.make_derived_metric(*, metric, transform, sample_param_names=None, allow_decrypt=False)

Encrypted-aware analogue of :func:fairlearn.metrics.make_derived_metric.

A metric outside the encrypted catalogue raises ValueError on encrypted input unless allow_decrypt=True, which decrypts y_pred (emitting DecryptFallbackWarning).