Computer Science

Evaluating explainability for graph neural networks

C. Agarwal, O. Queen, et al.

As the use of Graph Neural Networks (GNNs) expands in critical applications, evaluating the quality and reliability of their explanations becomes vital. This paper introduces SHAPEGGEN, a versatile synthetic graph data generator that produces benchmark datasets with ground-truth explanations, paving the way for rigorous assessments. The research was conducted by Chirag Agarwal, Owen Queen, Himabindu Lakkaraju, and Marinka Zitnik.