Research map / Graph learning / Graph self-supervised learning
Graph Information Bottleneck for Subgraph Recognition
Junchi Yu, Tingyang Xu, Yu Rong, Yatao Bian, Junzhou Huang, Ran He
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 |
|---|---|---|---|
| Apply GIB to GIN | GIN · MUTAG | accuracy: 0.825$\pm$0.068 -> 0.839 $\pm$ 0.064 | No clear effect |
| Apply GIB to GIN | GIN · PROTEINS | accuracy: 0.707 $\pm$ 0.056 -> 0.749 $\pm$ 0.051 | No clear effect |
| Apply GIB to GIN | GIN · IMDB-BINARY | accuracy: 0.732 $\pm$ 0.048 -> 0.737 $\pm$ 0.070 | No clear effect |
| Apply GIB to GraphSAGE | GraphSAGE · DD | accuracy: 0.729 $7$ 0.041 -> 0.781 $\pm$ 0.042 | Improved |
| Recognize an IB-subgraph with GIB | GCN · MUTAG with 30% redundant edges | accuracy: 0.667 -> 0.684 | Improved |
| Recognize an IB-subgraph with GIB | GCN · MUTAG with 30% redundant edges | Recall: 0.324$\pm$ 0.049 -> 0.493$\pm$ 0.035 | Improved |
| Recognize an IB-subgraph with GIB | GCN · MUTAG with 30% redundant edges | Precision: 0.675$\pm$ 0.104 -> 0.692 $\pm$0.061 | No clear effect |
| Add mutual-information loss (ablation) | GCN+GIB · ZINC QED | absolute property bias: 0.43$\pm$ 0.15 -> 0.38$\pm$ 0.12 | Improved |
| Add mutual-information loss (ablation) | GCN+GIB · ZINC DRD2 | absolute property bias: 0.21$\pm$ 0.13 -> 0.06$\pm$ 0.09 | Improved |
| Add mutual-information loss (ablation) | GCN+GIB · ZINC HLM-CLint | absolute property bias: 0.48$\pm$ 0.34 -> 0.37$\pm$ 0.30 | Improved |
| Add mutual-information loss (ablation) | GCN+GIB · ZINC MLM-CLint | absolute property bias: 1.20$\pm$ 0.97 -> 0.72$\pm$ 0.55 | Improved |
| Add connectivity loss (ablation) | GCN+GIB · ZINC MLM-CLint | absolute property bias: 1.58$\pm$ 0.86 -> 0.72$\pm$ 0.55 | Improved |
Closest papers by meaning
- Recognizing Predictive Substructures With Subgraph Information Bottleneck (TPAMI 2021 · similarity 0.95)
- Improving Subgraph Recognition With Variational Graph Information Bottleneck (CVPR 2022 · similarity 0.94)
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.88)
- Implicit Subgraph Neural Network (ICML 2025 · similarity 0.86)
- Union Subgraph Neural Networks (AAAI 2024 · similarity 0.86)
- Translating Subgraphs to Nodes Makes Simple GNNs Strong and Efficient for Subgraph Representation Learning (ICML 2024 · similarity 0.86)
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.
