This is the official Pytorch implementation of our paper PerFedMask: Personalized Federated Learning with Optimized Masking Vectors accepted in ICLR 2023.

First check the requirements as follows:
python=3.7
numpy=1.17.0
pytorch=1.12.1
cudatoolkit = 11.3.1
wandb=0.12.19
torchvision=0.13.1
cvxpy=1.1.11
mosek=9.2.40

Then clone the repository as follows:

git clone https://github.com/MehdiSet/PerFedMask.git

We conduct our experiments on CIFAR-10, CIFAR-100, and DomainNet datasets using ResNet (PreResNet18), MobileNet , and AlexNet, respectively. Please download the datasets and place them under data/ directory.

If you find our paper and code useful, please cite our paper as follows:

@inproceedings{setayesh2023perfedmask,
  title={PerFedMask: {Personalized} Federated Learning with Optimized Masking Vectors},
  author={Setayesh, Mehdi and Li, Xiaoxiao and W.S. Wong, Vincent},
  booktitle={Proc. of International Conference on Learning Representations (ICLR)},
  address={Kigali, Rwanda},
  month={May},
  year={2023}
}

Please feel free to contact us if you have any questions:

• Mehdi Setayesh: setayeshm@ece.ubc.ca

This codebase was adapted from https://github.com/illidanlab/SplitMix and https://github.com/jhoon-oh/FedBABU.