GraphXAI
GraphXAI is an open-source library for evaluating graph neural network (GNN) explainers. This library provides XAI-ready datasets, a number of state-of-the-art explainers, and evaluation metrics for explanations. One of the major features of this package is the novel, flexible ShapeGGEN dataset generator which can generate graph datasets with unique ground-truth explanations. In addition, ShapeGGEN is parameterized such that generated graphs can have varying sizes, degree distributions, types of ground-truth explanations, levels of homophily/heterophily, and degrees of fairness as defined by a protected feature.
Installation
After cloning the repo, install the graphxai package from the root directory of this project:
pip install -e .
This will allow you to access features within the package, including datasets, explainers, and evaluation tools.
Data Availability
Downloads are provided for the datasets in our package through our page on the Harvard Dataverse.
Citation
Please use the following BibTeX to cite this project in your work:
@inproceedings{agarwal2022evaluating,
title = {Evaluating Explainability for Graph Neural Networks},
author = {Agarwal, Chirag and Queen, Owen and Lakkaraju, Himabindu and Zitnik, Marinka},
booktitle = {arXiv:2208.09339},
year = {2022}
}