Towards Comprehensive and Prerequisite-Free Explainer for Graph Neural Networks
Han Zhang, Yan Wang, Guanfeng Liu, Pengfei Ding, Huaxiong Wang, Kwok-Yan Lam
TL;DR
OPEN introduces a prerequisite-free explainer for GNNs that jointly infers environments spanning the dataset's sample space and learns invariant, environment-aware subgraphs to reveal the full decision logic of GNNs under distribution shifts. Using NPAF to partition data into $K$ environments and GVAG to sample subgraphs with node/graph-invariant embeddings, the framework avoids requiring access to GNN internals or edge-feature constraints. The method combines NodeVAE, NGD-based subgraph generation, and a suite of regularizations and loss terms to optimize explanations that align with predictions across environments, achieving superior fidelity and robustness in both prerequisite-free and prerequisite-satisfied scenarios. Experiments on GOOD with $Cora$ and $Motif$ demonstrate substantial improvements in fidelity metrics and efficiency, particularly on complex distributions, while ablation confirms the necessity of environment inference and invariant learning components. This work advances XGNN by enabling comprehensive, practical, and scalable explanations for GNN decisions in real-world, distribution-shifting contexts.
Abstract
To enhance the reliability and credibility of graph neural networks (GNNs) and improve the transparency of their decision logic, a new field of explainability of GNNs (XGNN) has emerged. However, two major limitations severely degrade the performance and hinder the generalizability of existing XGNN methods: they (a) fail to capture the complete decision logic of GNNs across diverse distributions in the entire dataset's sample space, and (b) impose strict prerequisites on edge properties and GNN internal accessibility. To address these limitations, we propose OPEN, a novel c\textbf{O}mprehensive and \textbf{P}rerequisite-free \textbf{E}xplainer for G\textbf{N}Ns. OPEN, as the first work in the literature, can infer and partition the entire dataset's sample space into multiple environments, each containing graphs that follow a distinct distribution. OPEN further learns the decision logic of GNNs across different distributions by sampling subgraphs from each environment and analyzing their predictions, thus eliminating the need for strict prerequisites. Experimental results demonstrate that OPEN captures nearly complete decision logic of GNNs, outperforms state-of-the-art methods in fidelity while maintaining similar efficiency, and enhances robustness in real-world scenarios.