Research map / Graph learning / GNN architectures
Sign and Basis Invariant Networks for Spectral Graph Representation Learning
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 |
|---|---|---|---|
| Use SignNet positional encodings instead of the strongest listed baseline | Graph-level molecular regression · ZINC (10K) | test MAE: .122\std{.01} -> .084\std{.006} | Helped |
| Use SignNet positional encodings instead of the strongest listed baseline | Graph-level molecular regression · ZINC-full | test MAE: .042\std{.003} -> .024\std{.003} | Helped |
| Use SignNet positional encodings instead of the strongest listed baseline | Graph-level molecular regression · Alchemy (10k) | test MAE: .111\std{.01} -> .113\std{.002} | Hurt |
| Use all available eigenvectors instead of 8 (ablation) | SignNet with GatedGCN · ZINC | test MAE: $0.121_{\pm 0.005}$ -> $0.100_{\pm 0.007}$ | Helped |
| Use all available eigenvectors instead of a fixed number (ablation) | SignNet with PNA · ZINC | test MAE: $0.105_{\pm 0.007}$ -> $0.084_{\pm 0.006}$ | Helped |
| Add sign-invariant processing instead of using φ(v) only (ablation) | SignNet with GatedGCN · ZINC | test MAE: $0.148_{\pm 0.007}$ -> $0.121_{\pm 0.005}$ | Helped |
| Use SignNet instead of the original intrinsic neural field | Texture reconstruction · Cat mesh | LPIPS: .189 -> .147 | Helped |
| Use SignNet instead of the original intrinsic neural field | Texture reconstruction · Human mesh | DSSIM: .119 -> .125 | Hurt |
Closest papers by meaning
- On the Expressive Power of Spectral Invariant Graph Neural Networks (ICML 2024 · similarity 0.90)
- On the Generalization of Equivariant Graph Neural Networks (ICML 2024 · similarity 0.89)
- Spectral Graph Neural Networks are Incomplete on Graphs with a Simple Spectrum (NeurIPS 2025 · similarity 0.89)
- Position: Spectral GNNs Rely Less on Graph Fourier Basis than Conceived (ICML 2025 · similarity 0.88)
- On the Stability of Expressive Positional Encodings for Graphs (ICLR 2024 · similarity 0.88)
- Weisfeiler-Leman Is Incomplete on Simple Spectrum Graphs, so Canonicalize Them (NeurIPS 2026 · 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.