Exploring Explainability Methods for Graph Neural Networks

TL;DR

This paper evaluates various explainability methods applied to Graph Attention Networks (GATs) in graph-based image classification, providing insights into their effectiveness across multiple datasets.

Contribution

It demonstrates the applicability and assesses the performance of popular explainability techniques specifically on GATs for image classification tasks.

Findings

Explainability methods vary in effectiveness across datasets

Qualitative and quantitative assessments reveal strengths and limitations

Insights contribute to understanding GNN interpretability

Abstract

With the growing use of deep learning methods, particularly graph neural networks, which encode intricate interconnectedness information, for a variety of real tasks, there is a necessity for explainability in such settings. In this paper, we demonstrate the applicability of popular explainability approaches on Graph Attention Networks (GAT) for a graph-based super-pixel image classification task. We assess the qualitative and quantitative performance of these techniques on three different datasets and describe our findings. The results shed a fresh light on the notion of explainability in GNNs, particularly GATs.