Research map / Graph learning / Graph self-supervised learning
Self-supervised Graph-level Representation Learning with Local and Global Structure
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 |
|---|---|---|---|
| Add hierarchical global-semantic learning to the local graph and subgraph objectives (ablation) | GraphLoG with GIN · Biological function prediction | ROC-AUC (%): 71.3±0.7 -> 72.9±0.7 | Helped |
| Add the graph-level local objective to subgraph and global objectives (ablation) | GraphLoG with GIN · Biological function prediction | ROC-AUC (%): 71.9±0.8 -> 72.9±0.7 | Helped |
| Add the subgraph-level local objective to graph and global objectives (ablation) | GraphLoG with GIN · Biological function prediction | ROC-AUC (%): 72.2±0.4 -> 72.9±0.7 | Helped |
| Combine graph-level and subgraph-level local objectives instead of using either alone (ablation) | GraphLoG with GIN · Biological function prediction | ROC-AUC (%): 70.1±0.6; 71.0±0.3 -> 71.3±0.7 | Helped |
| Use global-semantic learning alone rather than local objectives alone (ablation) | GraphLoG with GIN · Biological function prediction | ROC-AUC (%): 70.1±0.6; 71.0±0.3 -> 71.5±0.5 | Helped |
| Pre-train with GraphLoG rather than GraphCL | GIN · MoleculeNet eight-task average | ROC-AUC (%): 71.3 -> 73.4 | Helped |
| Pre-train with GraphLoG rather than GraphCL | GIN · Biological function prediction | ROC-AUC (%): 71.2±0.6 -> 72.9±0.7 | Helped |
| Use GraphLoG pre-training rather than the strongest self-supervised baseline on each task | GIN · MoleculeNet (BBBP, Tox21, ToxCast, SIDER, ClinTox, MUV, HIV, BACE) | ROC-AUC (%) | Mixed |
Closest papers by meaning
- InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization (ICLR 2020 · similarity 0.90)
- Pre-Training Graph Neural Networks on Molecules by Using Subgraph-Conditioned Graph Information Bottleneck (AAAI 2025 · similarity 0.87)
- Robust Heterogeneous Graph Classification for Molecular Property Prediction with Information Bottleneck (AAAI 2025 · similarity 0.87)
- Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks (ICML 2025 · similarity 0.86)
- GraphGPT: Generative Pre-trained Graph Eulerian Transformer (ICML 2025 · similarity 0.86)
- Enhancing Molecular Representation Learning Through the Combination of 3D and 2D Graph Machine Learning (Student Abstract) (AAAI 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.