Federated Learning With Opacus (PyTorch)

In this repository, I used PyTorch and Opacus to simulate the fedavg and fedprox algorithm of federated learning. And add differential privacy in the models.

Requirments

• python == 3.8.0
• pytorch == 1.12.1
• torchvision == 0.13.1
• opacus == 1.4.0
• numpy == 1.23.2
• matplotlib == 3.2.0
• scikit-learn == 1.2.2

Running the experiments

• Running fedavg algorithm with differential privacy

python main.py --dp True

• Running fedprox algorithm with differential privacy

python main.py --dp True --fedprox True

Advantage

Compared to the fedavg done by others, my model is more featured, you can see my code.