Research map / Graph learning / GNN architectures
Expressive Graph Neural Networks via Equivariant Use of Noise
Xiyuan Wang, Muhan Zhang
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 equivariant noise (ENGNN-O) (ablation) | MPNN · Squirrel | accuracy: 51.99 -> 64.66 | Improved |
| add equivariant noise (ENGNN-O) (ablation) | MPNN · Actor | accuracy: 40.41 -> 45.76 | Improved |
| add equivariant noise (ENGNN-O) (ablation) | MPNN · DDI link prediction | HR@20: 24.76 -> 77.61 | Improved |
| add equivariant noise (ENGNN-P) (ablation) | MPNN · DDI link prediction | HR@20: 24.76 -> 27.64 | No clear effect |
| add inequivariant noise (NMPNN) (ablation) | MPNN · Squirrel | accuracy: 51.99 -> 38.25 | Hurt |
| add inequivariant noise (NMPNN) (ablation) | MPNN · Actor | accuracy: 40.41 -> 23.73 | Hurt |
| add inequivariant noise (NMPNN) (ablation) | MPNN · ZINC | MAE: 0.131 -> 0.136 | Hurt |
| use equivariant noise instead of inequivariant noise (ENGNN-O vs NMPNN) (ablation) | NMPNN · ZINC | MAE: 0.136 -> 0.070 | Improved |
| add noise (ENGNN-P) to MPNN for counting (ablation) | MPNN · 6-cycle counting | MAE: 0.1313 -> 0.0078 | Improved |
| use equivariant instead of inequivariant noise (ENGNN-P vs NMPNN) (ablation) | NMPNN · 6-cycle counting | MAE: 0.0228 -> 0.0078 | Improved |
| add equivariant noise (ENGNN-O) (ablation) | MPNN · ogbg-molhiv | ROC-AUC: 78.27 -> 78.63 | No clear effect |
| replace MPNN with ENGNN-O for scalability vs PPGN | PPGN · ZINC GPU memory | memory (GB): 20.37 -> 0.62 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BernNet — BernNet: Learning Arbitrary Graph Spectral Filters via Bernstein Approximation (NeurIPS 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN Expressiveness (ICLR 2024 · similarity 0.92)
- Towards Bridging Generalization and Expressivity of Graph Neural Networks (ICLR 2025 · similarity 0.92)
- Schur Nets: exploiting local structure for equivariance in higher order graph neural networks (NeurIPS 2024 · similarity 0.90)
- On the Rademacher Complexity of Graph Neural Networks: Unifying Expressivity and Geometry (NeurIPS 2026 · similarity 0.90)
- An Empirical Study of Realized GNN Expressiveness (ICML 2024 · similarity 0.90)
- Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks? (NeurIPS 2024 · similarity 0.90)
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.
