Introduction

This is the implementation of our paper Eliminating Domain Bias for Federated Learning in Representation Space (accepted by NeurIPS 2023). It can improve bi-directional knowledge transfer between the server and clients. We show the code of the representative FedAvg+DBE (FedDBE).

• [NeurIPS] DBE.pptx
• DBEPoster.pdf

Takeaway: By eliminating domain bias in the feature extractor, we address catastrophic forgetting during local training, enhancing the generalization ability. Consequently, the global module can swiftly adapt to a new client.

Citation

@inproceedings{zhang2023eliminating,
  title={Eliminating Domain Bias for Federated Learning in Representation Space},
  author={Jianqing Zhang and Yang Hua and Jian Cao and Hao Wang and Tao Song and Zhengui XUE and Ruhui Ma and Haibing Guan},
  booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
  year={2023},
  url={https://openreview.net/forum?id=nO5i1XdUS0}
}

Dataset

Due to the file size limitation, we only upload the fmnist dataset with the default practical setting ($\beta=0.1$). Please refer to our project PFLlib.

System (based on PFL-Non-IID)

• main.py: configurations of FedDBE.
• run_me.sh: command lines to start FedDBE.
• env_linux.yaml: python environment to run FedDBE on Linux.
• ./flcore:
  • ./clients/clientDBE.py: the code on the client.
  • ./servers/serverDBE.py: the code on the server.
  • ./trainmodel/models.py: the code for models.
• ./utils:
  • data_utils.py: the code to read the dataset.

Simulation

Environments

With the installed conda, we can run this platform in a conda virtual environment called fl.

conda env create -f env_cuda_latest.yaml # for Linux

Training and Evaluation

All codes corresponding to FedDBE are stored in ./system. Just run the following commands.

cd ./system
sh run_me.sh