Research map / Graph learning / Graph self-supervised learning
Improving Subgraph Recognition With Variational Graph Information Bottleneck
Junchi Yu, Jie Cao, Ran He
Check 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 |
|---|---|---|---|
| Replace GIB mutual-information estimation with VGIB noise injection and its variational objective | GCN subgraph recognition · QED, HLM-CLint, MLM-CLint, and RLM-CLint | absolute property divergence: QED: 0.38±0.12 -> 0.32±0.12; HLM-CLint: 0.37±0.30 -> 0.34±0.28; MLM-CLint: 0.72±0.55 -> 0.69±0.58; RLM-CLint: 1.15±0.68 | Improved |
| Use VGIB instead of GIB | GCN subgraph recognition · QED | training time: 1712.93s -> 146.53s | Improved |
| Use VGIB to generate explanations instead of the strongest listed baseline | GCN explanations · RLM-CLint | Fidelity+: 0.694 -> 0.765 | Improved |
| Use VGIB to generate explanations instead of the strongest listed baseline | GCN explanations · HLM-CLint | Fidelity+: 0.782 -> 0.792 | Improved |
| Use VGIB to generate explanations instead of the strongest listed baseline | GCN explanations · QED | Fidelity+: 0.602 -> 0.627 | Improved |
| Use VGIB to generate explanations instead of the strongest listed baseline | GCN explanations · DRD2 | Fidelity+: 0.74 -> 0.756 | Improved |
| Use VGIB rather than the strongest listed baseline | GCN explanations · RLM-CLint | Fidelity-: 0.478 -> 0.463 | Improved |
| Use VGIB rather than the strongest listed baseline | GCN explanations · HLM-CLint | Fidelity-: 0.616 -> 0.579 | Improved |
| Use VGIB rather than the strongest listed baseline | GCN explanations · QED | Fidelity-: 0.498 -> 0.487 | Improved |
| Use VGIB rather than the strongest listed baseline | GCN explanations · DRD2 | Fidelity-: 0.428 -> 0.424 | Improved |
| Add VGIB subgraph recognition to GIN | GIN with sum pooling · MUTAG | accuracy: 0.84±0.06 -> 0.86±0.08 | Improved |
| Add VGIB subgraph recognition to GIN | GIN with sum pooling · PROTEINS | accuracy: 0.74±0.05 -> 0.75±0.04 | Improved |
Closest papers by meaning
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.88)
- Subgraph Invariant Learning Towards Large-Scale Graph Node Classification (AAAI 2025 · similarity 0.87)
- Translating Subgraphs to Nodes Makes Simple GNNs Strong and Efficient for Subgraph Representation Learning (ICML 2024 · similarity 0.86)
- An Efficient Subgraph-Inferring Framework for Large-Scale Heterogeneous Graphs (AAAI 2024 · similarity 0.86)
- Union Subgraph Neural Networks (AAAI 2024 · similarity 0.86)
- Generalizing Weisfeiler-Lehman Kernels to Subgraphs (ICLR 2025 · 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.
