/soteriafl

Code for paper SoteriaFL: A Unified Framework for Private Federated Learning with Communication Compression

Primary LanguageJupyter Notebook

Numerical experiments for SoteriaFL

This repo contains numerical experiments code for "SoteriaFL: A Unified Framework for Private Federated Learning with Communication Compression" [PDF].

Files:

  • optimizers.py: optimization algorithms.

  • SoteriaFL_a9a.ipynb: numerical experiments on the a9a dataset.

  • SoteriaFL_NN.ipynb: numerical experiments on the MNIST dataset.

Dependency: Please install [this package]

If you find this repo useful, please cite our paper

@article{li2022soteriafl,
  title={{SoteriaFL}: A unified framework for private federated learning with communication compression},
  author={Li, Zhize and Zhao, Haoyu and Li, Boyue and Chi, Yuejie},
  journal={Advances in Neural Information Processing Systems},
  volume = {35},
  year={2022}
}