Statistical Test for Saliency Maps of Graph Neural Networks via Selective Inference

Shuichi Nishino, Tomohiro Shiraishi, Teruyuki Katsuoka, Ichiro Takeuchi

Abstract: Graph Neural Networks (GNNs) have gained prominence for their ability to process graph-structured data across various domains. However, interpreting GNN decisions remains a significant challenge, leading to the adoption of saliency maps for identifying salient subgraphs composed of influential nodes and edges. Despite their utility, the reliability of GNN saliency maps has been questioned, particularly in terms of their robustness to input noise. In this study, we propose a statistical testing framework to rigorously evaluate the significance of saliency maps. Our main contribution lies in addressing the inflation of the Type I error rate caused by double-dipping of data, leveraging the framework of Selective Inference. Our method provides statistically valid $p$-values while controlling the Type I error rate, ensuring that identified salient subgraphs contain meaningful information rather than random artifacts. The method is applicable to a variety of saliency methods with piecewise linearity (e.g., Class Activation Mapping). We validate our method on synthetic and real-world datasets, demonstrating its capability in assessing the reliability of GNN interpretations.

Submission Length: Regular submission (no more than 12 pages of main content)

Changes Since Last Submission:

(1) In response to the AE's additional comments, we revised the paper to clarify the relationship between statistical significance and interpretability.

- In Section 1, we added a concise statement in the related works:
"Statistical significance alone is not sufficient for interpretability. Our method can be viewed as a complementary metric of interpretability; see Section 5."

- In Section 5, we fully revised the paragraph on Interpretability to discuss this point in detail.

(2) Other changes
- De-anonymization
- Added acknowledgments
- Added Code URL (removed Supplementary Material)
- Slightly adjusted figure/table sizes and expressions to improve the layout

Code: https://github.com/ni-shu/si_for_gnn

Assigned Action Editor: ~Christopher_Morris1

Submission Number: 4924