Explaining GNN Explanations with Edge Gradients

2025-08-01

AI Summary

Existing GNN interpretability methods exhibit inconsistent performance across complex architectures and tasks, lacking a unified theoretical foundation. To address this, we establish, for the first time, a rigorous theoretical connection between perturbation-based and gradient-based methods, introducing *edge gradients* as a unifying explanatory tool: their sign approximates GNNExplainer’s output, and in linear GNNs, they are equivalent to edge masking search. Our framework integrates edge gradient computation, perturbation-based evaluation, gradient backpropagation, and computational graph subgraph extraction. We systematically validate it on both synthetic and real-world graph datasets. Experiments demonstrate that edge gradients significantly improve explanation reliability and cross-model consistency—offering a novel interpretability paradigm for GNNs that balances theoretical rigor with practical effectiveness.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Computer Vision: Interpretability, Explainability, and Transparency
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

In recent years, the remarkable success of graph neural networks (GNNs) on graph-structured data has prompted a surge of methods for explaining GNN predictions. However, the state-of-the-art for GNN explainability remains in flux. Different comparisons find mixed results for different methods, with many explainers struggling on more complex GNN architectures and tasks. This presents an urgent need for a more careful theoretical analysis of competing GNN explanation methods. In this work we take a closer look at GNN explanations in two different settings: input-level explanations, which produce explanatory subgraphs of the input graph, and layerwise explanations, which produce explanatory subgraphs of the computation graph. We establish the first theoretical connections between the popular perturbation-based and classical gradient-based methods, as well as point out connections between other recently proposed methods. At the input level, we demonstrate conditions under which GNNExplainer can be approximated by a simple heuristic based on the sign of the edge gradients. In the layerwise setting, we point out that edge gradients are equivalent to occlusion search for linear GNNs. Finally, we demonstrate how our theoretical results manifest in practice with experiments on both synthetic and real datasets.

Problem

• Analyzing theoretical connections between GNN explanation methods
• Comparing input-level and layerwise GNN explanation approaches
• Validating edge gradient-based heuristics for GNN explainability

Innovation

• Analyzing GNN explanations using edge gradients
• Connecting perturbation-based and gradient-based methods
• Validating theory with synthetic and real datasets