PFLlib: A Beginner-Friendly and Comprehensive Personalized Federated Learning Library and Benchmark

TL;DR

The paper tackles the challenge of tracking, implementing, and benchmarking personalized federated learning methods amid rapid advances. It introduces PFLlib, a beginner-friendly library that implements 37 SOTA algorithms (8 tFL, 29 pFL) and an integrated benchmark across diverse datasets and heterogeneity scenarios. The framework includes privacy evaluation via DLG attacks and PSNR, and a modular server-client design with simple APIs to add new algorithms and scenarios. By providing a standardized platform and extensive coverage, PFLlib aims to accelerate pFL research, enable fair comparisons, and lower the entry barrier for newcomers. The work demonstrates tangible community impact through widespread adoption and continued expansion of pFL capabilities.

Abstract

Amid the ongoing advancements in Federated Learning (FL), a machine learning paradigm that allows collaborative learning with data privacy protection, personalized FL (pFL)has gained significant prominence as a research direction within the FL domain. Whereas traditional FL (tFL) focuses on jointly learning a global model, pFL aims to balance each client's global and personalized goals in FL settings. To foster the pFL research community, we started and built PFLlib, a comprehensive pFL library with an integrated benchmark platform. In PFLlib, we implemented 37 state-of-the-art FL algorithms (8 tFL algorithms and 29 pFL algorithms) and provided various evaluation environments with three statistically heterogeneous scenarios and 24 datasets. At present, PFLlib has gained more than 1600 stars and 300 forks on GitHub.