PGM-Explainer: Probabilistic Graphical Model Explanations for Graph Neural Networks

TL;DR

PGM-Explainer introduces a probabilistic graphical model approach to explain GNN predictions by identifying key graph components and modeling feature dependencies, outperforming existing methods on benchmarks.

Contribution

It presents a novel PGM-based explainer for GNNs that captures feature dependencies and includes the Markov blanket, providing more comprehensive explanations.

Findings

Achieves better performance than existing explainers on benchmark tasks.

Effectively identifies crucial graph components for explanations.

Models feature dependencies through conditional probabilities.

Abstract

In Graph Neural Networks (GNNs), the graph structure is incorporated into the learning of node representations. This complex structure makes explaining GNNs' predictions become much more challenging. In this paper, we propose PGM-Explainer, a Probabilistic Graphical Model (PGM) model-agnostic explainer for GNNs. Given a prediction to be explained, PGM-Explainer identifies crucial graph components and generates an explanation in form of a PGM approximating that prediction. Different from existing explainers for GNNs where the explanations are drawn from a set of linear functions of explained features, PGM-Explainer is able to demonstrate the dependencies of explained features in form of conditional probabilities. Our theoretical analysis shows that the PGM generated by PGM-Explainer includes the Markov-blanket of the target prediction, i.e. including all its statistical information. We also show that the explanation returned by PGM-Explainer contains the same set of independence statements in the perfect map. Our experiments on both synthetic and real-world datasets show that PGM-Explainer achieves better performance than existing explainers in many benchmark tasks.