What is the FedLab Tool Assembly?
FedLab is an open-source federated learning framework developed by SMILE Lab. It is designed for flexibility and modularity, supporting a wide range of federated algorithms, data distributions, and research scenarios. FedLab provides comprehensive baselines, data partitioning tools, and supports both synchronous and asynchronous FL, making it suitable for research and simulation of realistic FL conditions.
What can you use the FedLab Tool Assembly for?
- Research in federated learning under realistic (IID and non-IID) conditions
- Simulation of diverse federated algorithms and data distributions
- Benchmarking and evaluation of FL systems
Limitations/Remarks
- Effectiveness in highly heterogeneous environments or under specific operational constraints may vary
- Primarily focused on research and simulation, not production
Specific Use Cases/Applications
- FL research with custom data partitions and algorithms
- Evaluation of FL robustness and performance
- Simulation of real-world FL scenarios
How to access the FedLab Tool Assembly?
- Official Documentation
- GitHub Repository
- Install via pip:
pip install fedlab
References
- Zeng, D. et al. (2021). FedLab: A Flexible Federated Learning Framework. arXiv:2107.11621. PDF
- FedLab Documentation
- FedLab GitHub
Original Framework Contributors
- Dun Zeng
- Siqi Liang
- Xiangjing Hu
- Zenglin Xu
- SMILE Lab
For a full list, see the FedLab GitHub contributors.
More information
Training
Affiliations Contributors
Editor
KU Leuven / University of UHasselt