Research map / Graph learning / GNN architectures
Learning on Random Balls is Sufficient for Estimating (Some) Graph Parameters
Takanori Maehara, Hoang NT
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 |
|---|---|---|---|
| observe only random balls (55.7% of data) for train and test instead of full graphs | GIN-style GNN · COLLAB | test accuracy (%): 80.2 -> 80.3 | No clear effect |
| observe only random balls (14.9% memory) instead of full graph | GIN-style GNN · RDT-BINARY | test accuracy (%): 92.4 -> 79.0 | Hurt |
| observe only random balls (14.2% memory) instead of full graph | GIN-style GNN · RDT-MULTI5K | test accuracy (%): 57.5 -> 44.0 | Hurt |
| RBS-GNN with partial observation vs WL subtree kernel | RBS-GNN[≡2] · COLLAB | test accuracy (%): 78.9 -> 80.3 | Improved |
| RBS-GNN with partial observation vs WL subtree kernel | RBS-GNN[≡2] · RDT-BINARY | test accuracy (%): 81.0 -> 79.0 | Hurt |
| RBS-GNN with partial observation vs PatchySan | RBS-GNN[≡2] · COLLAB | test accuracy (%): 72.6 -> 80.3 | Improved |
| RBS-GNN with partial observation vs AWL | RBS-GNN[≡2] · COLLAB | test accuracy (%): 73.9 -> 80.3 | Improved |
| RBS-GNN with partial observation vs AWL | RBS-GNN[≡2] · RDT-MULTI5K | test accuracy (%): 54.7 -> 44.0 | Hurt |
| size generalization target: non-estimable max-degree instead of estimable global clustering coefficient | RBS-GNN[≡2] · COLLAB | test accuracy (%): 23.7 -> 16.1 | Hurt |
| size generalization target: non-estimable max-degree instead of estimable global clustering coefficient | RBS-GNN[≡2] · RDT-BINARY | test accuracy (%): 33.7 -> 43.4 | Improved |
| size generalization target: non-estimable max-degree instead of estimable global clustering coefficient | RBS-GNN[≡2] · RDT-MULTI5K | test accuracy (%): 40.0 -> 39.8 | No clear effect |
Closest papers by meaning
- Graph Convolution for Semi-Supervised Classification: Improved Linear Separability and Out-of-Distribution Generalization (ICML 2021 · similarity 0.87)
- Learning Graph Representation via Graph Entropy Maximization (ICML 2024 · similarity 0.87)
- Almost Surely Asymptotically Constant Graph Neural Networks (NeurIPS 2024 · similarity 0.87)
- Graph Classification via Reference Distribution Learning: Theory and Practice (NeurIPS 2024 · similarity 0.86)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.86)
- Robust Offline Active Learning on Graphs (NeurIPS 2024 · 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.
