COMBINEX: A Unified Counterfactual Explainer for Graph Neural Networks via Node Feature and Structural Perturbations

2025-02-14

AI Summary

Existing GNN explanation methods predominantly focus on edge-structure perturbations, overlooking the critical role of node-feature perturbations. This work proposes COMBINEX—the first counterfactual explanation framework that jointly models perturbations to both node features and graph structure, targeting node and graph classification tasks. It generates minimal, realistic, and interpretable explanations that flip model predictions. Its key innovations are: (1) joint optimization of continuous/discrete node-feature and edge-structure perturbations; and (2) integration of differentiable counterfactual optimization, gradient-guided search, and multi-objective minimization—ensuring compatibility with diverse GNN architectures. Extensive evaluation on multiple real-world graph datasets demonstrates that COMBINEX significantly outperforms baselines: counterfactual validity improves by over 23%, and average perturbation magnitude decreases by 37%.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Search and Optimization: Combinatorial Optimization
• 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

Counterfactual explanations have emerged as a powerful tool to unveil the opaque decision-making processes of graph neural networks (GNNs). However, existing techniques primarily focus on edge modifications, often overlooking the crucial role of node feature perturbations in shaping model predictions. To address this limitation, we propose COMBINEX, a novel GNN explainer that generates counterfactual explanations for both node and graph classification tasks. Unlike prior methods, which treat structural and feature-based changes independently, COMBINEX optimally balances modifications to edges and node features by jointly optimizing these perturbations. This unified approach ensures minimal yet effective changes required to flip a model's prediction, resulting in realistic and interpretable counterfactuals. Additionally, COMBINEX seamlessly handles both continuous and discrete node features, enhancing its versatility across diverse datasets and GNN architectures. Extensive experiments on real-world datasets and various GNN architectures demonstrate the effectiveness and robustness of our approach over existing baselines.

Problem

• Unveils GNN decision-making via counterfactual explanations
• Balances node feature and structural perturbations
• Handles continuous and discrete node features

Innovation

• Unifies node feature and structural perturbations
• Optimally balances edge and node modifications
• Handles continuous and discrete node features