Discrete Diffusion-Based Model-Level Explanation of Heterogeneous GNNs with Node Features

TL;DR

This paper introduces DiGNNExplainer, a diffusion-based method for generating realistic, faithful explanations of heterogeneous GNNs at the model level, effectively handling discrete node features.

Contribution

It presents a novel diffusion-based approach for explaining HGNNs that supports realistic discrete node features, improving interpretability over existing methods.

Findings

Outperforms state-of-the-art explanation methods

Produces realistic discrete node features

Provides faithful explanations of model decisions

Abstract

Many real-world datasets, such as citation networks, social networks, and molecular structures, are naturally represented as heterogeneous graphs, where nodes belong to different types and have additional features. For example, in a citation network, nodes representing "Paper" or "Author" may include attributes like keywords or affiliations. A critical machine learning task on these graphs is node classification, which is useful for applications such as fake news detection, corporate risk assessment, and molecular property prediction. Although Heterogeneous Graph Neural Networks (HGNNs) perform well in these contexts, their predictions remain opaque. Existing post-hoc explanation methods lack support for actual node features beyond one-hot encoding of node type and often fail to generate realistic, faithful explanations. To address these gaps, we propose DiGNNExplainer, a model-level explanation approach that synthesizes heterogeneous graphs with realistic node features via discrete denoising diffusion. In particular, we generate realistic discrete features (e.g., bag-of-words features) using diffusion models within a discrete space, whereas previous approaches are limited to continuous spaces. We evaluate our approach on multiple datasets and show that DiGNNExplainer produces explanations that are realistic and faithful to the model's decision-making, outperforming state-of-the-art methods.