Generating Robust Counterfactual Witnesses for Graph Neural Networks

TL;DR

This paper proposes robust counterfactual witnesses (RCWs) as a new explanation method for graph neural networks, ensuring explanations remain valid under graph perturbations, with theoretical hardness results and scalable algorithms demonstrated on benchmark datasets.

Contribution

Introduces the concept of RCWs for GNN explanations, providing algorithms for their verification and generation, along with theoretical hardness analysis and scalability solutions.

Findings

Efficient algorithms for RCW verification and generation.

Theoretical hardness results for RCW verification.

Experimental validation on benchmark datasets.

Abstract

This paper introduces a new class of explanation structures, called robust counterfactual witnesses (RCWs), to provide robust, both counterfactual and factual explanations for graph neural networks. Given a graph neural network M, a robust counterfactual witness refers to the fraction of a graph G that are counterfactual and factual explanation of the results of M over G, but also remains so for any "disturbed" G by flipping up to k of its node pairs. We establish the hardness results, from tractable results to co-NP-hardness, for verifying and generating robust counterfactual witnesses. We study such structures for GNN-based node classification, and present efficient algorithms to verify and generate RCWs. We also provide a parallel algorithm to verify and generate RCWs for large graphs with scalability guarantees. We experimentally verify our explanation generation process for benchmark datasets, and showcase their applications.