Research map / Graph learning / GNN architectures
Graph Neural Networks with Learnable and Optimal Polynomial Bases
Yuhe Guo, Zhewei Wei
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 |
|---|---|---|---|
| learn an optimal signal-specific polynomial basis | OptBasisGNN · YCbCr synthetic filter learning | MSE loss: 0.0058 vs 0.1501 (ChebII) | Improved |
| learn an arbitrary orthonormal basis rather than the optimal one (ablation) | polynomial filter learning · YCbCr synthetic filter learning | MSE loss: 0.0058 vs 0.3175 | Hurt |
| use a Monomial basis rather than the optimal basis | polynomial filter learning · YCbCr synthetic filter learning | MSE loss: 0.0058 vs 3.9076 | Hurt |
| apply the scaled-up OptBasisGNN with batch training | OptBasisGNN · Wiki | accuracy: 60.95 -> 61.85 | Improved |
| use learnable orthonormal polynomial bases | FavardGNN · Squirrel | accuracy: 57.72 -> 63.49 | Improved |
| use the optimal polynomial basis | OptBasisGNN · Chameleon | accuracy: 74.20 -> 74.26 | Improved |
| use the optimal polynomial basis | OptBasisGNN · ogbn-arxiv | accuracy: 72.32 -> 72.27 | No clear effect |
| train FavardGNN for extended epochs without early stopping (ablation) | FavardGNN · YCbCr synthetic filter learning | MSE loss | Hurt |
| use early stopping rather than continuing FavardGNN training (ablation) | FavardGNN · node classification and filter learning | loss | Improved |
Closest papers by meaning
- How Universal Polynomial Bases Enhance Spectral Graph Neural Networks: Heterophily, Over-smoothing, and Over-squashing (ICML 2024 · similarity 0.92)
- Position: Spectral GNNs Rely Less on Graph Fourier Basis than Conceived (ICML 2025 · similarity 0.90)
- Unifying Homophily and Heterophily for Spectral Graph Neural Networks via Triple Filter Ensembles (NeurIPS 2024 · similarity 0.89)
- SLOG: An Inductive Spectral Graph Neural Network Beyond Polynomial Filter (ICML 2024 · similarity 0.89)
- WaveNet: Tackling Non-stationary Graph Signals via Graph Spectral Wavelets (AAAI 2024 · similarity 0.87)
- Improving Expressive Power of Spectral Graph Neural Networks with Eigenvalue Correction (AAAI 2024 · similarity 0.87)
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.
