FairD-PFL

This repository contains the official code for our FairD-PFL model. The method combines personalized federated learning with fairness-aware regularization for multi-site neuroimaging classification. The code has been standardized and successfully passed preliminary tests. It is operational, though it may still requires a deeper technical review.

Highlights

• Dual-view brain connectivity features (AAL + CC200)
• Personalized federated training with fairness regularization
• Default evaluation with subject-level K-fold cross validation

Quick Start

1. Prepare an HDF5 dataset in the same format as our experiments.
2. Update the path in args or pass --hdf5_path on the command line.
3. Run training:

python main.py

Notes

• This codebase is provided for research and reproducibility only.
• See LICENSE for usage restrictions before formal publication of FairD-PFL.