Research map / Graph learning / GNN architectures
From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness
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 |
|---|---|---|---|
| Apply GNN-AK+ to GIN | GIN · ZINC-12K | MAE: 0.163 ± 0.004 -> 0.080 ± 0.001 | Helped |
| Apply GNN-AK+ to GIN | GIN · CIFAR10 | accuracy: 59.82 ± 0.33 -> 72.19 ± 0.13 | Helped |
| Apply GNN-AK+ to GIN | GIN · PATTERN | accuracy: 85.732 ± 0.023 -> 86.850 ± 0.057 | Helped |
| Use 1-hop egonets instead of the full method's 3-hop egonets (ablation) | GIN-AK+ · ZINC-12K | MAE: 0.080 ± 0.001 -> 0.147 ± 0.006 | Hurt |
| Use 2-hop egonets instead of the full method's 3-hop egonets (ablation) | GIN-AK+ · ZINC-12K | MAE: 0.080 ± 0.001 -> 0.120 ± 0.005 | Hurt |
| Use 1-hop egonets instead of 2-hop egonets (ablation) | GIN-AK+ · CIFAR10 | accuracy: 72.19 ± 0.13 -> 71.37 ± 0.28 | Hurt |
| Use 3-hop egonets instead of 2-hop egonets (ablation) | GIN-AK+ · CIFAR10 | accuracy: 72.19 ± 0.13 -> OOM | Mixed |
| Add subgraph encoding (ablation) | GIN-AK+ · ZINC-12K | MAE: 0.086 ± 0.001 -> 0.080 ± 0.001 | Helped |
| Add subgraph encoding (ablation) | GIN-AK+ · CIFAR10 | accuracy: 67.76 ± 0.29 -> 72.19 ± 0.13 | Helped |
| Add centroid encoding (ablation) | GIN-AK+ · ZINC-12K | MAE: 0.084 ± 0.003 -> 0.080 ± 0.001 | Helped |
| Add centroid encoding (ablation) | GIN-AK+ · CIFAR10 | accuracy: 72.20 ± 0.96 -> 72.19 ± 0.13 | No clear effect |
| Add context encoding (ablation) | GIN-AK+ · ZINC-12K | MAE: 0.088 ± 0.003 -> 0.080 ± 0.001 | Helped |
Closest papers by meaning
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.91)
- Subgraphormer: Unifying Subgraph GNNs and Graph Transformers via Graph Products (ICML 2024 · similarity 0.90)
- Translating Subgraphs to Nodes Makes Simple GNNs Strong and Efficient for Subgraph Representation Learning (ICML 2024 · similarity 0.90)
- Generative Graph Pattern Machine (NeurIPS 2025 · similarity 0.89)
- Probabilistic Graph Rewiring via Virtual Nodes (NeurIPS 2024 · similarity 0.89)
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.89)
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.