Research map / Graph learning / GNN architectures
On Graph Neural Networks versus Graph-Augmented MLPs
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 |
|---|---|---|---|
| replace GA-MLP-A with GIN | attributed-walk counting on Cora · Cora | MSE loss divided by label variance: 1.56E-1 -> 9.72E-7 | Helped |
| replace GA-MLP-A with GIN | attributed-walk counting on RRG · RRG | MSE loss divided by label variance: 2.13E-2 -> 2.61E-4 | Helped |
| use twice as many powers of the operator (ablation) | GA-MLP-A on Cora attributed-walk counting · Cora train | MSE loss divided by label variance: 1.23E-1 -> 1.87E-2 | Helped |
| use twice as many powers of the operator (ablation) | GA-MLP-A on Cora attributed-walk counting · Cora test | MSE loss divided by label variance: 1.56E-1 -> 6.44E-2 | Helped |
| use twice as many powers of the operator (ablation) | GA-MLP-A on RRG attributed-walk counting · RRG | MSE loss divided by label variance: 1.75E-2 -> 1.69E-2; 2.13E-2 -> 2.13E-2 | Mixed |
| replace A with the Bethe Hessian operator family | GA-MLP for community detection · binary SBM | overlap | Helped |
| replace A with the normalized operator family Ã_(1) (ablation) | GA-MLP for community detection · binary SBM | overlap | No clear effect |
| use an A-only sGNN with the same receptive field | GA-MLP-A comparison for community detection · binary SBM | overlap | No clear effect |
| use GA-MLP-A instead of GNN | rooted-graph equivalence class separation at K=2 · Cora, Citeseer, and Pubmed rooted graphs | number of equivalence classes: 1589 vs 756; 984 vs 506; 8059 vs 3762 | Hurt |
Closest papers by meaning
- Graph Neural Networks are Inherently Good Generalizers: Insights by Bridging GNNs and MLPs (ICLR 2023 · similarity 0.91)
- A Theoretical Comparison of Graph Neural Network Extensions (ICML 2022 · similarity 0.91)
- A New Perspective on "How Graph Neural Networks Go Beyond Weisfeiler-Lehman?" (ICLR 2022 · similarity 0.91)
- Analyzing the Expressive Power of Graph Neural Networks in a Spectral Perspective (ICLR 2021 · similarity 0.91)
- $\mathscr{N}$-WL: A New Hierarchy of Expressivity for Graph Neural Networks (ICLR 2023 · similarity 0.90)
- Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness (ICLR 2024 · similarity 0.89)
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.