Mitigating Bias in Federated Learning

TL;DR

This paper addresses bias in federated learning by proposing three privacy-preserving methods to reduce discrimination, analyzing their effectiveness across diverse data distributions and party participation levels.

Contribution

It introduces three novel bias mitigation techniques specifically designed for federated learning, maintaining data privacy while improving fairness.

Findings

Methods are effective with skewed data distributions.

Significant bias reduction when at least 20% of parties use the methods.

Improved fairness metrics without compromising model accuracy.

Abstract

As methods to create discrimination-aware models develop, they focus on centralized ML, leaving federated learning (FL) unexplored. FL is a rising approach for collaborative ML, in which an aggregator orchestrates multiple parties to train a global model without sharing their training data. In this paper, we discuss causes of bias in FL and propose three pre-processing and in-processing methods to mitigate bias, without compromising data privacy, a key FL requirement. As data heterogeneity among parties is one of the challenging characteristics of FL, we conduct experiments over several data distributions to analyze their effects on model performance, fairness metrics, and bias learning patterns. We conduct a comprehensive analysis of our proposed techniques, the results demonstrating that these methods are effective even when parties have skewed data distributions or as little as 20% of parties employ the methods.