arpitdm/gnn_accuracy_fairness_tradeoff: v0.1.0-alpha
No Thumbnail Available
Restricted Availability
Date
2023-08-16, 2023-08-16
Persistent identifier of the Data Catalogue metadata
Creator/contributor
Editor
Journal title
Journal volume
Publisher
Publication Type
software
Peer Review Status
Repositories
Access rights
ISBN
ISSN
Description
Reference code and data for analyzing the tradeoff between accuracy and algorithmic fairness on graph neural networks for node classification. This repository provides support for downloading and preprocessing four datasets (German, Credit, Penn94, and Pokec-Z), building and training three GNN models (GCN, GraphSAGE, and GIN), and implementing algorithmic fairness interventions including PFR-AX and PostProcess (ours), Unaware, EDITS, and NIFTY (baselines).