DRO-FAIR: Distributionally Robust Optimization for joint Demographic Parity + Individual Fairness under adversarial data corruption (PGD/FGSM attacks, coordinated label flips). Implements Algorithm 1 from ICML submission with 150 experiments across Adult, Credit, and LSAC datasets.
machine-learning pytorch fairness dro demographic-parity adversarial-ml individual-fairness distributionally-robust-optimization fairml
-
Updated
Aug 11, 2026 - Python