Research map / Graph learning / GNN architectures
A graph similarity for deep learning
Seongmin Ok
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 aggregate-transform aggregations (GAT/GCN/GraphSAGE/WWL) with transform-sum-cat | WLS kernel · DHFR | accuracy: 67.74±4.00 -> 77.92±4.78 | Improved |
| replace aggregate-transform aggregations with transform-sum-cat | WLS kernel · Synthie | accuracy: 46.00±3.69 -> 86.79±5.82 | Improved |
| replace aggregate-transform aggregations with transform-sum-cat | WLS kernel · COX2 | accuracy: 79.26±3.54 -> 77.95±3.48 | Hurt |
| add nonlinear feature map transform (vs identity map WLSLin) (ablation) | WLS kernel · Synthie | accuracy: 55.00±7.22 -> 86.79±5.82 | Improved |
| add feature map transform (vs identity map WLSLin) (ablation) | WLS kernel · DHFR | accuracy: 68.26±2.59 -> 77.92±4.78 | Improved |
| use WLS GNN instead of GIN, GatedGCN etc. | WLS GNN · SBM_PATTERN | accuracy: 99.03±0.04 -> 99.08±0.02 | Improved |
| use WLS GNN instead of GatedGCN | WLS GNN · SBM_CLUSTER | accuracy: 60.20±0.48 -> 60.49±0.41 | Improved |
| use WLS GNN instead of best baseline | WLS GNN · ZINC | MAE: 0.362±0.001 -> 0.332±0.007 | Improved |
| add edge attributes to WLS GNN (ablation) | WLS GNN · ZINC | MAE: 0.332±0.007 -> 0.315±0.003 | Improved |
| use WLS GNN vs other isotropic GNNs | WLS GNN · TU graph classification | accuracy: 63.07 vs 75.85 (PROTEINS); 70.42 vs 68.58 (ENZYMES) | Mixed |
| replace R-GCN discriminator with WLS discriminator | MolGAN · molecule generation validity | valid molecules % | Improved |
Closest papers by meaning
- How Powerful are Graph Neural Networks? (ICLR 2019 · similarity 0.88)
- Learning Graph Neural Networks with Approximate Gradient Descent (AAAI 2021 · similarity 0.87)
- Deep Learning on Graphs: A Data-Centric Exploration (AAAI 2024 · similarity 0.86)
- On the equivalence between graph isomorphism testing and function approximation with GNNs (NeurIPS 2019 · similarity 0.86)
- Exploring Consistency in Graph Representations: from Graph Kernels to Graph Neural Networks (NeurIPS 2024 · similarity 0.85)
- What do Graph Neural Networks learn? Insights from Tropical Geometry (NeurIPS 2024 · similarity 0.84)
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.
