mlsanityData to 14 Oct 2026

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

arXiv 2505.16893 ↗First posted 22 May 2025stat.ML cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 32 of its 48 references that are in the map. First posted on arXiv 22 May 2025.

Builds on

32 of 48 references.

References · 32

  1. 2017Graph Attention Networkscited by 8,923
  2. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  3. 2018How Powerful are Graph Neural Networks?cited by 4,393
  4. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  5. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  6. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  7. 2019GNNExplainer: Generating Explanations for Graph Neural Networkscited by 737
  8. 2019Graph Transformer Networkscited by 523
  9. 2020Explainability in Graph Neural Networks: A Taxonomic Surveycited by 384
  10. 2021On Explainability of Graph Neural Networks via Subgraph Explorationscited by 281
  11. 2020GraphLIME: Local Interpretable Model Explanations for Graph Neural Networkscited by 209
  12. 2013Exact post-selection inference, with application to the lassocited by 91
20 more
  1. 2014Optimal Inference After Model Selectioncited by 67
  2. 2021GraphSVX: Shapley Value Explanations for Graph Neural Networkscited by 64
  3. 2014Exact Post-Selection Inference for Sequential Regression Procedurescited by 48
  4. 2021Zorro: Valid, Sparse, and Stable Explanations in Graph Neural Networkscited by 35
  5. 2021EEG-GNN: Graph Neural Networks for Classification of Electroencephalogram (EEG) Signalscited by 28
  6. 2021More Powerful Conditional Selective Inference for Generalized Lasso by Parametric Programmingcited by 25
  7. 2019Computing Valid p-values for Image Segmentation by Selective Inferencecited by 24
  8. 2020Quantifying Statistical Significance of Neural Network-based Image Segmentation by Selective Inferencecited by 21
  9. 2023Graph Neural Network-based EEG Classification: A Surveycited by 21
  10. 2024Graph Neural Network Explanations are Fragilecited by 19
  11. 2014Selecting the number of principal components: estimation of the true rank of a noisy matrixcited by 18
  12. 2014A significance test for forward stepwise model selectioncited by 17
  13. 2023Valid P-Value for Deep Learning-Driven Salient Regioncited by 17
  14. 2021Tree-Values: selective inference for regression treescited by 15
  15. 2017Unifying approach to selective inference with applications to cross-validationcited by 9
  16. 2024Statistical Test for Anomaly Detections by Variational Auto-Encoderscited by 7
  17. 2024Explainable Graph Neural Networks Under Firecited by 5
  18. 2025si4onnx: A Python package for Selective Inference in Deep Learning Modelscited by 5
  19. 2024Statistical Test for Feature Selection Pipelines by Selective Inferencecited by 3
  20. 2024Statistical Test for Auto Feature Engineering by Selective Inferencecited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Statistical Testing for Multiple Instance Learning via Selective Inference with Applications to Computational Pathologycited by 0