Research map / Graph learning / GNN architectures
A New Perspective on the Effects of Spectrum in Graph Neural Networks
Mingqi Yang, Yanming Shen, Rui Li, Heng Qi, Qiang Zhang, Baocai Yin
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 the correlation-free independent-filter architecture rather than shared filters (ablation) | GNN graph convolution · ZINC | test MAE: 0.1415±0.00748 -> 0.0944±0.00379; 0.1061±0.01018 -> 0.0698±0.00200 | Improved |
| Use Aρ rather than normalized symmetric adjacency as the filter basis (ablation) | Correlation-free GNN · ZINC | test MAE: 0.0944±0.00379 -> 0.0698±0.00200 | Improved |
| Use generalized normalization rather than normalized symmetric adjacency as the filter basis (ablation) | Correlation-free GNN · ZINC | test MAE: 0.0944±0.00379 -> 0.0709±0.00176 | Improved |
| Increase polynomial basis count (ablation) | Correlation-free GNN with Aρ · ZINC | MAE | Improved |
| Increase model depth (ablation) | Correlation-free GNN · ZINC | MAE | No clear effect |
| Increase model depth beyond 20 layers (ablation) | Shared-filter GNN · ZINC | MAE | Hurt |
| Use Aρ instead of normalized symmetric adjacency (ablation) | GNN with shared filters · ZINC | test MAE: 0.1415±0.00748 -> 0.1061±0.01018 | Improved |
| Use generalized normalization instead of normalized symmetric adjacency (ablation) | GNN with shared filters · ZINC | test MAE: 0.1415±0.00748 -> 0.1133±0.01711 | Improved |
| Use Aρ rather than generalized normalization (ablation) | Correlation-free GNN · ZINC | test MAE: 0.0709±0.00176 -> 0.0698±0.00200 | Improved |
Closest papers by meaning
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.91)
- Position: Spectral GNNs Rely Less on Graph Fourier Basis than Conceived (ICML 2025 · similarity 0.90)
- On the Expressive Power of Spectral Invariant Graph Neural Networks (ICML 2024 · similarity 0.90)
- SLOG: An Inductive Spectral Graph Neural Network Beyond Polynomial Filter (ICML 2024 · similarity 0.89)
- Unifying Homophily and Heterophily for Spectral Graph Neural Networks via Triple Filter Ensembles (NeurIPS 2024 · similarity 0.89)
- HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations (NeurIPS 2025 · similarity 0.88)
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.
