Research map / Graph learning / GNN architectures
Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries
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 |
|---|---|---|---|
| use SUN layers (separate root parameters, local+global pooling) instead of GNN-AK+ in a subgraph GNN | Subgraph GNN with EGO+ policy · Counting Substructures: Star | MAE: 0.0150 -> 0.0064 | Helped |
| use SUN (EGO+) instead of GNN-AK+ | GIN-based Subgraph GNN · Counting Substructures: Triangle | MAE: 0.0123 -> 0.0079 | Helped |
| use SUN (EGO+) instead of PPGN (3-WL model) | 4-Cycle counting · Counting Substructures: 4-Cycle | MAE: 0.0090 -> 0.0105 | Hurt |
| use SUN (EGO) instead of GNN-AK+ | GIN-based Subgraph GNN · ZINC-12k | MAE: 0.091 -> 0.083 | Helped |
| use SUN (EGO+) instead of DSS-GNN (EGO+) | GIN-based Subgraph GNN · ogbg-molhiv | ROC-AUC (%): 76.78 -> 80.03 | Helped |
| use SUN (EGO+) instead of GNN-AK+ | GIN-based Subgraph GNN · ogbg-molhiv | ROC-AUC (%): 79.61 -> 80.03 | Helped |
| use SUN instead of CIN (ring-aware model) | ogbg-molhiv · ogbg-molhiv | ROC-AUC (%): 80.94 -> 80.03 | No clear effect |
| use SUN instead of DSS-, DS-GNN and GNN-AK variants with limited training data | Subgraph GNNs · ZINC-12k low-data generalisation | MAE | Helped |
| use SUN instead of other Subgraph GNNs with limited training data | Subgraph GNNs · 4-cycle counting low-data generalisation | MAE | Mixed |
Closest papers by meaning
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.94)
- Subgraphormer: Unifying Subgraph GNNs and Graph Transformers via Graph Products (ICML 2024 · similarity 0.92)
- Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality (ICML 2025 · similarity 0.92)
- Translating Subgraphs to Nodes Makes Simple GNNs Strong and Efficient for Subgraph Representation Learning (ICML 2024 · similarity 0.91)
- Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness (ICLR 2024 · similarity 0.91)
- Efficient Subgraph GNNs by Learning Effective Selection Policies (ICLR 2024 · similarity 0.90)
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.