Research map / Graph learning / GNN architectures
How to Find Your Friendly Neighborhood: Graph Attention Design with Self-Supervision
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 SuperGAT-MX instead of GAT | node classification models · 17 real-world datasets | accuracy | Mixed |
| increase the edge self-supervision loss weight λE | SuperGAT-GO and SuperGAT-DP · Cora, CiteSeer, PubMed, PPI | link prediction AUC | Helped |
| increase the edge self-supervision loss weight λE | SuperGAT-GO and SuperGAT-DP · Cora, CiteSeer, PubMed, PPI | node classification accuracy or micro F1 | Hurt |
| use DP attention instead of GO attention | graph attention for link prediction · Cora, CiteSeer, PubMed | AUC | Helped |
| use GO attention instead of DP attention | graph attention for label agreement · Cora and other citation datasets | KL divergence | Helped |
| use scaled dot-product attention instead of GAT | SuperGAT-SD · synthetic graphs with homophily ≤ 0.2 | accuracy | Helped |
| use mixed GO and DP attention instead of scaled dot-product attention | SuperGAT-MX · synthetic graphs with homophily above 0.2 and moderate average degree | accuracy | Helped |
| use SuperGAT-MX instead of SuperGAT-SD (ablation) | SuperGAT · Four-Univ | accuracy: 76.2 ± 0.8 -> 75.3 ± 0.6 | Hurt |
| use SuperGAT-SD instead of SuperGAT-MX (ablation) | SuperGAT · Cora | accuracy: 84.3 ± 0.6 -> 82.7 ± 0.6 | Hurt |
| use SuperGAT-MX instead of SuperGAT-SD (ablation) | SuperGAT · PPI | micro F1: 74.4 ± 0.4 -> 67.2 ± 1.2 | Hurt |
| add edge self-supervision to graph attention | SuperGAT · 15 of 17 real-world datasets | node classification accuracy | Helped |
Closest papers by meaning
- Understanding Attention and Generalization in Graph Neural Networks (NeurIPS 2019 · similarity 0.90)
- Tailoring Self-Attention for Graph via Rooted Subtrees (NeurIPS 2023 · similarity 0.89)
- Graph Attention is Not Always Beneficial: A Theoretical Analysis of Graph Attention Mechanisms via Contextual Stochastic Block Models (ICML 2025 · similarity 0.88)
- What Improves the Generalization of Graph Transformers? A Theoretical Dive into the Self-attention and Positional Encoding (ICML 2024 · similarity 0.87)
- HONGAT: Graph Attention Networks in the Presence of High-Order Neighbors (AAAI 2024 · similarity 0.87)
- Edge Prompt Tuning for Graph Neural Networks (ICLR 2025 · similarity 0.86)
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.