Simon007-heiyewuxing / StoCFL

Code for paper - Stochastic Clustered Federated Learning

Stochastic-Clustered-Federated-Learning

Prepare

pip install -r requirements.txt

Run

Stochastic federated client clusetering procedure (FCC)

python StoCFL-FCC.py --com_round 100 \
                     --num_per_round 20 \
                     --batch_size 50 \
                     --lr 0.1 \
                     --epochs 20 \
                     --mu 0.1 \
                     --tau 0.2 \
                     --setting fskewed \
                     --k 4 \
                     --process 1

Bi-level clustered federated learning (CFL)

python StoCFL-FCC+CFL.py --n 200 \
                         --com_round 200 \
                         --num_per_round 20 \
                         --batch_size 50 \
                         --lr 0.1 \
                         --epochs 5 \
                         --obbs 50 \ 
                         --obep 20 \   
                         --mu 0.05 \
                         --tau 0.65 \
                         --seed 0 \
                         --dataset cifar \
                         --process 1 

Citation

If you find this repo helpful to your research, please cite our paper:

@article{zeng2023stochastic,
  title={Stochastic clustered federated learning},
  author={Zeng, Dun and Hu, Xiangjing and Liu, Shiyu and Yu, Yue and Wang, Qifan and Xu, Zenglin},
  journal={arXiv preprint arXiv:2303.00897},
  year={2023}
}