Research map / Graph learning / Graph algorithms and clustering
Universal Function Approximation on Graphs
Rickard Brüel Gabrielsson
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 |
|---|---|---|---|
| Use NPA instead of GNN (GIN) | Graph classification · GNN-Hard | train accuracy: 50 -> 100 | Improved |
| Use NPA instead of GNN (GIN) | Graph classification · Random-Regular | train accuracy: 10 -> 90 | Improved |
| Add subgraph dropout to NPA (ablation) | NPA · NCI1 | validation accuracy: 81.8$\pm$1.9 -> 84.0$\pm$2.2 | Improved |
| Add subgraph dropout to NPA (ablation) | NPA · PTC | validation accuracy: 67.6$\pm$5.9 -> 67.1$\pm$6.9 | Hurt |
| Sort nodes by subgraph levels (ablation) | NPA · MUTAG | validation accuracy: 92.8$\pm$7.0 -> 93.3$\pm$6.0 | Improved |
| Sort nodes by subgraph levels (ablation) | NPA · PTC | validation accuracy: 67.6$\pm$5.9 -> 65.9$\pm$8.3 | Hurt |
| Sort edges by two endpoint degrees rather than not sorting (ablation) | NPA · NCI1 | validation accuracy: 77.7$\pm$3.0 -> 84.0$\pm$2.2 | Improved |
| Sort edges by two endpoint degrees rather than not sorting (ablation) | NPA · PROTEINS | validation accuracy: 76.9$\pm$3.0 -> 76.2$\pm$4.6 | No clear effect |
| Sort edges by degrees and labels rather than not sorting (ablation) | NPA · MUTAG | validation accuracy: 92.8$\pm$7.5 -> 88.9$\pm$10.5 | Hurt |
| Sort edges by two endpoint degrees rather than not sorting (ablation) | NPA · PTC | validation accuracy: 65.3$\pm$5.9 -> 67.6$\pm$5.9 | Improved |
| Use iso-injective NPA instead of non-iso-injective NPBA (ablation) | Graph classification · GNN-Hard | train accuracy: 50 -> 100 | Improved |
| Use iso-injective NPA instead of non-iso-injective NPBA (ablation) | Graph classification · NPBA-Hard | train accuracy: 50 -> 100 | Improved |
Closest papers by meaning
- Function graph transformers universally approximate operators between function spaces (NeurIPS 2026 · similarity 0.85)
- Primphormer: Efficient Graph Transformers with Primal Representations (ICML 2025 · similarity 0.85)
- Weisfeiler-Leman Is Incomplete on Simple Spectrum Graphs, so Canonicalize Them (NeurIPS 2026 · similarity 0.84)
- Weisfeiler and Leman Go Loopy: A New Hierarchy for Graph Representational Learning (NeurIPS 2024 · similarity 0.84)
- On topological descriptors for graph products (NeurIPS 2025 · similarity 0.84)
- Generalization, Expressivity, and Universality of Graph Neural Networks on Attributed Graphs (ICLR 2025 · 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.
