FedLab: A Versatile Framework for Federated Learning
Dun Zeng, Siqi Liang, Xiangjing Hu, Hui Wang, Zenglin Xu; 24(100):1−7, 2023.
Abstract
FedLab is an open-source framework that offers a lightweight and flexible solution for simulating federated learning. The main focus of FedLab’s design is to enhance the effectiveness of federated learning algorithms and improve communication efficiency. It provides customization options for server optimization, client optimization, communication agreements, and communication compression. Additionally, FedLab is scalable and can be used in various deployment scenarios with different computation and communication resources. Our goal is for FedLab to offer researchers in the federated learning community flexible APIs, reliable baseline implementations, and alleviate the challenges of implementing novel approaches. For the source code, tutorial, and documentation, please visit https://github.com/SMILELab-FL/FedLab.
[abs]