A Post-Processing-Based Fair Federated Learning Framework
Yi Zhou, Naman Goel
TL;DR
This paper tackles fairness in Federated Learning by decoupling global model training from fairness enforcement, enabling fully local debiasing after standard FL rounds. It introduces two post-processing-based schemes—model output post-processing and final-layer fairness fine-tuning—that let clients tailor fairness to their local data and requirements with little global communication cost. Across tabular, ECG, and imaging datasets under varying data heterogeneity, the framework consistently improves Equalized Odds with minimal or even positive effects on accuracy, especially in highly heterogeneous settings. The approach is simple, modular, and adaptable, offering a practical path to decentralized, privacy-preserving fairness in real-world FL deployments.
Abstract
Federated Learning (FL) allows collaborative model training among distributed parties without pooling local datasets at a central server. However, the distributed nature of FL poses challenges in training fair federated learning models. The existing techniques are often limited in offering fairness flexibility to clients and performance. We formally define and empirically analyze a simple and intuitive post-processing-based framework to improve group fairness in FL systems. This framework can be divided into two stages: a standard FL training stage followed by a completely decentralized local debiasing stage. In the first stage, a global model is trained without fairness constraints using a standard federated learning algorithm (e.g. FedAvg). In the second stage, each client applies fairness post-processing on the global model using their respective local dataset. This allows for customized fairness improvements based on clients' desired and context-guided fairness requirements. We demonstrate two well-established post-processing techniques in this framework: model output post-processing and final layer fine-tuning. We evaluate the framework against three common baselines on four different datasets, including tabular, signal, and image data, each with varying levels of data heterogeneity across clients. Our work shows that this framework not only simplifies fairness implementation in FL but also provides significant fairness improvements with minimal accuracy loss or even accuracy gain, across data modalities and machine learning methods, being especially effective in more heterogeneous settings.