fairGNN
an R package for training and evaluating fairness-aware Gated Neural Networks (GNN)
Embedded Files
View the public GitHub page:
The fairGNN R package is a cutting-edge suite of tools designed to train and analyse fairness-aware gated neural networks (GNN), empowering subgroup-sensitive prediction, transparent modelling, and interpretable insights across complex clinical datasets.
Delivering precision, equity, and clarity—right where healthcare decisions matter most. Methods draw on prior work in mixture-of-experts neural networks by Jordan and Jacobs (1994) and fairness-aware learning by Hardt, Price, and Srebro (2016).
fairGNN provides a complete pipeline for training and evaluating a Gated Neural Network (GNN) designed to mitigate demographic bias in predictive modelling. The package implements a fairness-aware GNN that uses a custom loss function to enforce the Equalized Odds fairness criterion by minimising the variance in True Positive and False Positive Rates across subgroups.This vignette demonstrates the full workflow using the GENDEP dataset to predict antidepressant response, focusing on fairness across gender subgroups.
Package link: CRAN: Package fairGNN
Abstract: Tools for training and analysing fairness-aware gated neural networks for subgroup-aware prediction and interpretation in clinical datasets. Methods draw on prior work in mixture-of-experts neural networks by Jordan and Jacobs (1994) <doi:10.1007/978-1-4471-2097-1_113>, fairness-aware learning by Hardt, Price, and Srebro (2016) <doi:10.48550/arXiv.1610.02413>, and personalised treatment prediction for depression by Iniesta, Stahl, and McGuffin (2016) <doi:10.1016/j.jpsychires.2016.03.016>.
Page updated
Google Sites
Report abuse