Research map / Graph learning / Graph self-supervised learning
Structural Entropy Guided Graph Hierarchical Pooling
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 fixed-quota, layerwise pooling with global structural-entropy-guided hierarchical assignments | SEP-G · IMDB-BINARY, IMDB-MULTI, COLLAB | accuracy: 73.48$\pm$0.76 -> 74.12$\pm$0.56; 51.31$\pm$0.72 -> 51.53$\pm$0.65; 80.87$\pm$0.34 -> 81.28$\pm$0.15 | Helped |
| replace fixed-quota, layerwise pooling with global structural-entropy-guided hierarchical assignments | SEP-G · MUTAG, PROTEINS | accuracy: 83.44$\pm$1.33 -> 85.56$\pm$1.09; 75.16$\pm$0.86 -> 76.42$\pm$0.39 | Helped |
| replace fixed-quota, layerwise pooling with global structural-entropy-guided hierarchical assignments | SEP-G · D&D, NCI1 | accuracy: 78.72$\pm$0.59 -> 77.98$\pm$0.57; 80.00$\pm$1.40 -> 78.35$\pm$0.33 | Mixed |
| use SEP-N hierarchical pooling and unpooling | SEP-N · Cora | accuracy: 84.4$\pm$0.6 -> 84.8$\pm$0.4 | Helped |
| use SEP-N hierarchical pooling and unpooling | SEP-N · Citeseer | accuracy: 73.6$\pm$0.09 -> 72.9$\pm$0.7 | Hurt |
| use SEP-N hierarchical pooling and unpooling | SEP-N · Pubmed | accuracy: 79.6$\pm$0.2 -> 80.2$\pm$0.8 | Helped |
| use 1 block instead of 2 (ablation) | SEP-N · Cora | accuracy: 84.8$\pm$0.4 -> 84.3$\pm$0.6 | Hurt |
| use 3, 4, or 5 blocks instead of 2 (ablation) | SEP-N · Cora | accuracy: 84.8$\pm$0.4 -> 84.5$\pm$0.3; 83.6$\pm$0.6; 83.9$\pm$0.5 | Hurt |
| use 1 block instead of 2 (ablation) | SEP-N · Citeseer | accuracy: 72.9$\pm$0.7 -> 73.3$\pm$0.6 | Helped |
| use 3, 4, or 5 blocks instead of 2 (ablation) | SEP-N · Citeseer | accuracy: 72.9$\pm$0.7 -> 72.1$\pm$0.6; 72.1$\pm$0.2; 72.4$\pm$0.6 | Hurt |
| use 1 block instead of 2 (ablation) | SEP-N · Pubmed | accuracy: 80.2$\pm$0.8 -> 78.9$\pm$0.6 | Hurt |
| use 3, 4, or 5 blocks instead of 2 (ablation) | SEP-N · Pubmed | accuracy: 80.2$\pm$0.8 -> 79.5$\pm$0.5; 78.5$\pm$0.3; 79.8$\pm$0.7 | Hurt |
Closest papers by meaning
- Geometry-Aware Edge Pooling for Graph Neural Networks (NeurIPS 2025 · similarity 0.89)
- SkipPool: Improved Sparse Hierarchical Graph Pooling with Differentiable Exploration (AAAI 2025 · similarity 0.89)
- Structural Entropy Based Graph Structure Learning for Node Classification (AAAI 2024 · similarity 0.88)
- Graph Parsing Networks (ICLR 2024 · similarity 0.88)
- Rhomboid Tiling for Geometric Graph Deep Learning (ICML 2025 · similarity 0.86)
- Enhancing Graph Classification Robustness with Singular Pooling (NeurIPS 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.