Research map / Graph learning / GNN architectures
Understanding Truncated Positional Encodings for Graph Neural Networks
James Flora, Mitchell Black, Weng-Keen Wong, Amir Nayyeri
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 |
|---|---|---|---|
| add eigenspace projections k=3 instead of k=2 (ablation) | k-EP-WL · BREC Total | accuracy (%): 32 -> 46.5 | Improved |
| increase k-EP-WL from k=3 to k=7 (ablation) | k-EP-WL · BREC Total | accuracy (%): 46.5 -> 48 | Improved |
| use single biharmonic distance instead of k-EP-WL k=7 | k-harmonics · BREC Total | accuracy (%): 48 -> 55 | Improved |
| use biharmonic instead of 4-harmonic (ablation) | k-harmonics · BREC Total | accuracy (%): 51 -> 55 | Improved |
| use biharmonic instead of resistance (ablation) | k-harmonics · BREC CFI | accuracy (%): 7 -> 11 | Improved |
| compare single k-harmonic with 3-WL on CFI graphs | k-harmonics · BREC CFI | accuracy (%): 60 -> 11 | Hurt |
| mix walks and k-harmonics (4+4) instead of k-harmonics alone (8) (ablation) | k-harmonic PE · ZINC-12k | test MAE: 0.076 -> 0.064 | Improved |
| mix projections and walks (4+4) instead of projections alone (8) (ablation) | projection PE · ZINC-12k | test MAE: 0.117 -> 0.075 | Improved |
| mix projections and walks (4+4) instead of walks alone (8) (ablation) | walk PE · ZINC-12k | test MAE: 0.082 -> 0.075 | Improved |
| mix projections and k-harmonics (4+4) instead of projections alone (8) (ablation) | projection PE · ZINC-12k | test MAE: 0.117 -> 0.078 | Improved |
| mix projections and k-harmonics (4+4) instead of k-harmonics alone (8) (ablation) | k-harmonic PE · ZINC-12k | test MAE: 0.076 -> 0.078 | No clear effect |
| mix walks and k-harmonics (4+4) instead of walks alone (8) (ablation) | walk PE · ZINC-12k | test MAE: 0.082 -> 0.064 | Improved |
Closest papers by meaning
- Learning Efficient Positional Encodings with Graph Neural Networks (ICLR 2025 · similarity 0.91)
- On the Stability of Expressive Positional Encodings for Graphs (ICLR 2024 · similarity 0.91)
- What Are Good Positional Encodings for Directed Graphs? (ICLR 2025 · similarity 0.90)
- On the Expressive Power of Spectral Invariant Graph Neural Networks (ICML 2024 · similarity 0.90)
- Homomorphism Expressivity of Spectral Invariant Graph Neural Networks (ICLR 2025 · similarity 0.89)
- 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.
