Research map / Graph learning / GNN architectures
Higher-Order Explanations of Graph Neural Networks via Relevant Walks
Thomas Schnake, Oliver Eberle, Jonas Lederer, Shinichi Nakajima, Kristof T. Schütt, Klaus‐Robert Müller, Grégoire Montavon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use LRP-based relevant-walk propagation instead of gradient-times-input (GNN-LRP vs GNN-GI) (ablation) | higher-order GNN explanation · Synthetic Spectral, activation task | AUFC: 0.87 -> 2.29 | Improved |
| use GNN-LRP instead of GNN-GI (ablation) | higher-order GNN explanation · VGG-16, pruning task | AUFC: 7.78 -> 6.05 | Improved |
| use GNN-LRP instead of GNN-GI (ablation) | higher-order GNN explanation · VGG-16, activation task | AUFC: 12.03 -> 14.04 | Improved |
| use relevant-walk explanation instead of first-order Pope LRP | GNN explanation · Synthetic Spectral, activation task | AUFC: 1.39 -> 2.29 | Improved |
| use GNN-LRP instead of GNNExplainer | GNN explanation · Synthetic GIN, pruning task | AUFC: 3.05 -> 2.85 | Improved |
| use GNN-LRP instead of GNNExplainer | GNN explanation · Sentiment GCN, activation task | AUFC: 20.23 -> 20.05 | No clear effect |
| use GNN-LRP instead of GNNExplainer | GNN explanation · SchNet-E, activation task | AUFC: 10.41 -> 10.47 | No clear effect |
| use GNN-LRP instead of GNNExplainer | GNN explanation · SchNet-mu, pruning task | AUFC: 0.28 -> 0.26 | Improved |
| use GNN-LRP instead of GNNExplainer | GNN explanation · SchNet-mu, activation task | AUFC: 1.01 -> 1.09 | Improved |
| use GNN-LRP instead of random explanation | GNN explanation · Synth-GIN, activation task | AUFC: 1.39 -> 3.92 | Improved |
Closest papers by meaning
- Higher Order Structures for Graph Explanations (AAAI 2025 · similarity 0.91)
- Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks (ICLR 2024 · similarity 0.91)
- RegExplainer: Generating Explanations for Graph Neural Networks in Regression Tasks (NeurIPS 2024 · similarity 0.90)
- Stratified GNN Explanations through Sufficient Expansion (AAAI 2024 · similarity 0.89)
- Generating In-Distribution Proxy Graphs for Explaining Graph Neural Networks (ICML 2024 · similarity 0.89)
- Factorized Explainer for Graph Neural Networks (AAAI 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
