Research map / Graph learning / GNN architectures
A PAC-Bayesian Approach to Generalization Bounds for Graph Neural Networks
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 the PAC-Bayes bound instead of the Rademacher complexity bound | MPGNN generalization analysis · MPGNN generalization bound | maximum node degree dependency: \mathcal{O}\left( d^{l-1} \sqrt{\log(d^{2l-3})} \right) -> \mathcal{O}\left( d^{l-1} \right) | Helped |
| Use the PAC-Bayes bound instead of the Rademacher complexity bound | MPGNN generalization analysis · MPGNN generalization bound | maximum hidden dimension dependency: \mathcal{O}\left( h \sqrt{\log h} \right) -> \mathcal{O}\left( \sqrt{h \log h} \right) | Helped |
| Use the proposed PAC-Bayes bound rather than the Rademacher complexity bound | MPGNN generalization analysis · Synthetic and real-world graph datasets | bound value | Helped |
| Increase message-passing steps/layers | MPGNN generalization-bound evaluation · Synthetic and real-world graph datasets | bound value | Hurt |
Closest papers by meaning
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.92)
- Towards Understanding Generalization of Graph Neural Networks (ICML 2023 · similarity 0.91)
- Compositional PAC-Bayes: Generalization of GNNs with persistence and beyond (NeurIPS 2024 · similarity 0.90)
- Generalization Error of Graph Neural Networks in the Mean-field Regime (ICML 2024 · similarity 0.89)
- Generalization Bounds via Meta-Learned Model Representations: PAC-Bayes and Sample Compression Hypernetworks (ICML 2025 · similarity 0.88)
- On the Rademacher Complexity of Graph Neural Networks: Unifying Expressivity and Geometry (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.