Research map / Graph learning / GNN architectures
Equivariant Subgraph Aggregation Networks
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 |
|---|---|---|---|
| Represent graphs as bags of edge-deleted subgraphs and encode them with DS-GNN using GIN | GIN · NCI109 | accuracy: 82.2$\pm$1.6 -> 83.0$\pm$1.7 | Helped |
| Represent graphs as bags of node-deleted subgraphs and encode them with DS-GNN using GIN | GIN · NCI1 | accuracy: 82.7$\pm$1.7 -> 83.8$\pm$2.4 | Helped |
| Represent graphs as bags of ego-networks and encode them with DS-GNN using GIN | GIN · ogbg-molhiv | ROC-AUC (%): 75.58$\pm$1.40 -> 78.00$\pm$1.42 | Helped |
| Represent graphs as bags of rooted ego-networks and encode them with DSS-GNN using GIN | GIN · ogbg-moltox21 | ROC-AUC (%): 74.91$\pm$0.51 -> 77.95$\pm$0.40 | Helped |
| Represent graphs as bags of rooted ego-networks and encode them with DSS-GNN using GIN | GIN · ZINC12k | MAE: 0.252$\pm$0.017 -> 0.102$\pm$0.003 | Helped |
| Add cross-subgraph information sharing (DSS-GNN instead of DS-GNN) (ablation) | ESAN with GIN · TUDatasets | accuracy | Mixed |
| Use stochastic subgraph sampling instead of the full subgraph set (ablation) | ESAN · TUDatasets | accuracy | Mixed |
| Use stochastic subgraph sampling at 5% of the subgraphs (ablation) | ESAN · EXP and CEXP | accuracy | Mixed |
| Use EGO or EGO+ rather than node deletion (ablation) | ESAN with GIN · ZINC12k | MAE: 0.171$\pm$0.010 -> 0.126$\pm$0.006 | Helped |
| Use EGO+ rather than edge deletion (ablation) | DSS-GNN with GIN · ZINC12k | MAE: 0.172$\pm$0.005 -> 0.102$\pm$0.003 | Helped |
| Add the original graph to the selected subgraph bag (ablation) | DS-GNN · TUDatasets | accuracy | Mixed |
| Use GIN instead of GCN as the base encoder (ablation) | ESAN · OGB datasets | ROC-AUC (%) | Mixed |
Closest papers by meaning
- A Flexible, Equivariant Framework for Subgraph GNNs via Graph Products and Graph Coarsening (NeurIPS 2024 · similarity 0.92)
- Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries (NeurIPS 2022 · similarity 0.91)
- Schur Nets: exploiting local structure for equivariance in higher order graph neural networks (NeurIPS 2024 · similarity 0.89)
- Subgraphormer: Unifying Subgraph GNNs and Graph Transformers via Graph Products (ICML 2024 · similarity 0.89)
- Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? (NeurIPS 2024 · similarity 0.89)
- Implicit Subgraph Neural Network (ICML 2025 · similarity 0.88)
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.