DeceFL: A Principled Decentralized Federated Learning Framework

This repository comprises codes that reproduce experiments in Ye, et al (2021), which has been submitted to Nature Machine Intelligence.

Project Organization

Experiments:

• Comparative studies of DeceFL and FedAvg, SL on dataset A2, provided in DatasetA2/.

• Time-vary experiments for DeceFL using dataset A2

  • Time-varying graphs with edge changes, provided in DatasetA2/
  • Time-varying graphs with node changes, provided in NodeVarying/

• Comparative study of DeceFL and FedAvg, SL on CWRU dataset, provided in CWRU/.

• An consensus example is generated by scripts in ConsensusExample/.

Go to each folder for README.md for every experiment.

Dependencies

Hardware: GPU that supports Pytorch

OS: Linux, Windows 10

Python packages:

• torch == 1.9.0
• numpy == 1.21.0
• sklearn == 0.24.2
• pandas == 1.3.1
• tqdm == 4.46.0
• matplotlib == 3.4.2

More to be filled ...

Reference

[1] Ye Yuan, et al. DeceFL: A Principled Decentralized Federated Learning Framework. Submitted for peer review, 2021.