Research map / Graph learning / GNN architectures
Predicting Global Label Relationship Matrix for Graph Neural Networks under Heterophily
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 LRGNN predicted label relationship matrix instead of OGNN, the runner-up | LRGNN · Squirrel | node classification accuracy (%): 62.44 -> 74.38 | Helped |
| Use LRGNN instead of GloGNN++ on Chameleon | LRGNN · Chameleon | node classification accuracy (%): 71.21 -> 79.16 | Helped |
| Use LRGNN instead of OGNN on Texas | LRGNN · Texas | node classification accuracy (%): 86.22 -> 90.27 | Helped |
| Use LRGNN instead of OGNN on Actor | LRGNN · Actor | node classification accuracy (%): 37.99 -> 37.34 | Hurt |
| Use LRGNN instead of LINKX on arXiv-year | LRGNN · arXiv-year | node classification accuracy (%): 56.00 -> 55.68 | Hurt |
| Use LRGNN instead of GloGNN++ on Penn94 | LRGNN · Penn94 | node classification accuracy (%): 85.74 -> 86.48 | Helped |
| Replace GCN with MLP (ignore graph) on Texas | MLP · Texas | node classification accuracy (%): 55.14 -> 80.81 | Helped |
| Use LRGNN instead of OGNN on Cora | LRGNN · Cora | node classification accuracy (%): 88.37 -> 88.33 | No clear effect |
| Vary operating rank q of the low-rank factorization | LRGNN · Wisconsin, Chameleon, Squirrel, Cornell | sign inference recovery error: 5 | Mixed |
| Degrade Texas features by adding Gaussian noise with larger sigma, compared with GloGNN++, H2GCN and MLP | LRGNN · Texas with noisy features | accuracy: 80% | Helped |
| Use low-rank prediction of label relationship matrix on synthetic graphs with low homophily | LRGNN · 15 synthetic random graphs | accuracy | Helped |
Closest papers by meaning
- Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs (NeurIPS 2020 · similarity 0.91)
- Combining Label Propagation and Simple Models out-performs Graph Neural Networks (ICLR 2021 · similarity 0.91)
- Graph Neural Networks with Heterophily (AAAI 2021 · similarity 0.90)
- On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks (NeurIPS 2024 · similarity 0.89)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.89)
- Integrating Co-Training with Edge Discrimination to Enhance Graph Neural Networks Under Heterophily (AAAI 2025 · similarity 0.89)
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.