Open Science Artifact - FedOverlap: A Weight Overlap Approach to Model Repair in Federated Learning

This package contains the codebase for FedOverlap and the results of the experiments reported in the paper "FedOverlap: A Federated Weight Overlap Approach to Model Repair in Federated Learning", which is currently under revision. The package contains the following file: README.md: Instructions to reproduce the entire study pipeline. Please refer to this file to run the experiments with FedOverlap and FedRepair; main.py: The script to run both FedOverlap and FedRepair approaches; sensitive-analysis-rq5.py: The script to run the sensitive analysis (RQ5) on FedOverlap; Preliminary_Analysis.ipynb: A Jupyter notebook for Section 5.1 (Preliminary Analysis), used to (i) analyze per-class performance after baseline federated training and (ii) select the target classes and experimental parameters for the repair study. Paper_Figures.ipynb: A Jupyter notebook file for pre-processing the results of the experiments and generating the plots of the paper; img/: Folder for saving the output plots and tables present in the paper; Paper Results/: Folder containing the .csv files with the results of the experiments.