Matching each group's normal data to a common target distribution via Sinkhorn distance yields an anomaly detector whose score distributions are approximately equal across groups, with better AUC-ADPD trade-offs than existing fair detectors.
How we analyzed the compas recidivism algorithm
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fairness-aware Anomaly Detection via Fair Projection
Matching each group's normal data to a common target distribution via Sinkhorn distance yields an anomaly detector whose score distributions are approximately equal across groups, with better AUC-ADPD trade-offs than existing fair detectors.