Skip to content

Latest commit

 

History

History
26 lines (22 loc) · 855 Bytes

README.md

File metadata and controls

26 lines (22 loc) · 855 Bytes
Raw

Confined Gradient Descent

Prerequisites

  1. Pytorch 1.9.0
  2. numpy 1.21.2
  3. scikit-learn 1.2.0
  4. pandas 1.3.2

Passive scripts

Unzip "pickled_mnist.pkl.zip" and run the dataset name.

  • mnist.py

Active scripts

  1. CGD experiment scripts: Start with cgd, followed by '_', then the dataset name.
  • cgd_cifar.py
  1. FedAvg experiment scripts: Start with fedavg, followed by '_', then the dataset name.
  • fedavg_cifar.py

Outputs

Collected in ./output, columns including epoch test_acc test_loss train_acc train_loss mia_acc idv_acc The first 5 columns are semantically self-explained. The mia_acc column shows the membership inference attack prediction accuracy (attack accuracy). The idv_acc shows the individual prediction accuracy for different CGD members. This column is not meaningful for FedAvg trainings.