Anh T. Do

Information Bottleneck Dominates Adversarial Training for Ancestry-Invariant Polygenic Risk Prediction: Dimensionality, Not Gradient Reversal, Controls the Fairness-Accuracy Tradeoff

This paper shows that latent dimensionality, not adversarial training, is the main factor controlling ancestry invariance in polygenic risk prediction, with dimensionality changes having a much larger impact on leakage than gradient reversal coefficients.