When majority rules, minority loses: bias amplification of gradient descent

François Bachoc, Jerome Bolte, Ryan Boustany, Jean-Michel Loubes

Keywords: Bias, fairness training, gradient descent, unbalanced learning

TL;DR: We analyze imbalanced training loss, showing that gradient descent dynamics can gradually reduce bias and recover minority-specific features with longer training.

Abstract: Despite growing empirical evidence of bias amplification in machine learning, its theoretical foundations remain poorly understood. We develop a formal framework for majority-minority learning tasks, showing how standard training can favor majority groups and produce stereotypical predictors that neglect minority-specific features. Assuming population and variance imbalance, our analysis reveals three key findings: (i) the close proximity between "full-data" and stereotypical predictors, (ii) the dominance of a region where training the entire model tends to merely learn the majority traits, and (iii) a lower bound on the additional training required. Our results are illustrated through experiments in deep learning for tabular and image classification tasks.

Primary Area: Optimization (e.g., convex and non-convex, stochastic, robust)

Submission Number: 13090