Research map / Graph learning / GNN architectures
Graph Neural Networks with 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 |
|---|---|---|---|
| Learn a class compatibility matrix and use it for propagation | CPGNN versus GCN-Cheby · syn-products, heterophily settings | accuracy: up to 7% | Helped |
| Use CPGNN-MLP-1 instead of MLP | syn-products · heterophily and homophily settings | accuracy: up to 30% better in heterophily and 50% better in homophily | Helped |
| Use CPGNN-Cheby-1 rather than GNN baselines | real heterophilous graphs with features · Texas, Squirrel, and Chameleon | mean accuracy: between 1.68% and 10.64% | Helped |
| Use compatibility-aware CPGNN variants | real homophilous graphs · CiteSeer, Pubmed, and Cora | accuracy | No clear effect |
| Initialize H using pretrained beliefs and known training labels rather than Glorot initialization (ablation) | CPGNN-MLP-1 · syn-products across homophily ratios | accuracy: up to 30% performance drop | Helped |
| Regularize H to keep its rows centered around zero (ablation) | CPGNN-MLP-1 · syn-products across homophily ratios | accuracy: up to 6% decrease in performance | Helped |
| Train H end-to-end after its initialization (ablation) | CPGNN-MLP-1 · syn-products across homophily ratios | accuracy: up to 21% performance gain | Helped |
| Pretrain the prior-belief estimator (ablation) | CPGNN-MLP-1 · syn-products, especially heterophily settings | accuracy: up to 2% increase | Helped |
Closest papers by meaning
- Is Homophily a Necessity for Graph Neural Networks? (ICLR 2022 · similarity 0.94)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.92)
- On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks (NeurIPS 2024 · similarity 0.92)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.91)
- Understanding Heterophily for Graph Neural Networks (ICML 2024 · similarity 0.91)
- Integrating Co-Training with Edge Discrimination to Enhance Graph Neural Networks Under Heterophily (AAAI 2025 · 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.