Users are allowed to integrate their own components, including datasets, models, etc., into FederatedScope to conduct federated learning for specific applications.
Fderated learning algorithms are modularized and expressed via defining events and corresponding handlers for the participants.
Developers can flexibly enrich the exchanged data and participants’ behaviors, which is helpful for various real-world federated learning applications.
We have implemented state-of-the-art personalized federated learning methods, and the well-designed interfaces make the development of new methods easy.
Out-of-the-box HPO functionalities can save users from the tedious loop of model tuning, allowing them to focus on their innovations.
Technologies, including differential privacy, encryption, multi-party computation, etc., are provided to enhance the strength of privacy protection.
02/2023, Our paper elaborating on FederatedScope is accepted at VLDB’23. pdf
10/2022, Our paper on personalized federated learning benchmark, pFL-Bench, is accepted at NeurIPS’22. pdf
08/2022, Our KDD 2022 paper on federated graph learning receives the KDD Best Paper Award for ADS track!
07/2022, We are hosting the CIKM’22 AnalytiCup competition. For more details, please see link
06/2022, We release pFL-Bench, a comprehensive benchmark for personalized Federated Learning (pFL), containing 10+ datasets and 20+ baselines. GitHub, pdf
06/2022, We release FedHPO-B, a benchmark suite for studying federated hyperparameter optimization. GitHub, pdf
06/2022, We release B-FHTL, a benchmark suit for studying federated hetero-task learning. GitHub, pdf
05/2022, Our paper on federated graph learning package is accepted at KDD’22. pdf
05/2022, FederatedScope v0.1.0, our first release, is available at GitHub.