Research map / Graph learning / GNN architectures
Interpretable Prototype-based Graph Information Bottleneck
Sangwoo Seo, Sungwon Kim, Chanyoung Park
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 |
|---|---|---|---|
| merge similar prototypes (ablation) | PGIB_cont · ZINC RLM | Fidelity F-: 0.441 -> 0.415 | Improved |
| merge similar prototypes (ablation) | PGIB_cont · ZINC HLM-CLint | Fidelity F-: 0.593 -> 0.543 | Improved |
| merge similar prototypes (ablation) | PGIB_cont · ZINC HLM-CLint | Fidelity F+: 0.772 -> 0.796 | Improved |
| merge similar prototypes (ablation) | PGIB_cont · ZINC DRD2 | Fidelity F+: 0.771 -> 0.781 | Improved |
| contrastive objective instead of variational IB | PGIB · MUTAG | accuracy: 85.00 -> 85.50 | Improved |
| contrastive objective instead of variational IB | PGIB · PROTEINS | accuracy: 77.14 -> 77.50 | Improved |
| mean-pooling readout instead of max-pooling (ablation) | PGIB_cont · MUTAG | accuracy: 85.50 -> 88.50 | Improved |
| mean-pooling readout instead of max-pooling (ablation) | PGIB_cont · PROTEINS | accuracy: 77.50 -> 68.39 | Hurt |
| sum-pooling readout instead of max-pooling (ablation) | PGIB_cont · NCI1 | accuracy: 78.25 -> 61.82 | Hurt |
| PGIB_cont instead of VGIB | VGIB · NCI1 | accuracy: 63.75 -> 78.25 | Improved |
| remove mutual information terms I(G;Gsub) or I(Gsub;Gp) (ablation) | PGIB · graph classification | accuracy | Hurt |
| remove prototype merging (ablation) | PGIB and PGIB_cont · graph classification | accuracy | Hurt |
Closest papers by meaning
- Factorized Explainer for Graph Neural Networks (AAAI 2024 · similarity 0.89)
- Self-Interpretable Graph Learning with Sufficient and Necessary Explanations (AAAI 2024 · similarity 0.88)
- Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks (ICLR 2024 · similarity 0.88)
- Generating In-Distribution Proxy Graphs for Explaining Graph Neural Networks (ICML 2024 · similarity 0.88)
- From GNNs to Trees: Multi-Granular Interpretability for Graph Neural Networks (ICLR 2025 · similarity 0.87)
- Stratified GNN Explanations through Sufficient Expansion (AAAI 2024 · similarity 0.87)
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.
