Research map / Graph learning / GNN architectures
Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs
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 |
|---|---|---|---|
| Separate ego and neighbor embeddings (ablation) | H2GCN-1 · syn-products, h<0.7 | accuracy: up to 40% | Helped |
| Include the ego-embedding alongside 1-hop and 2-hop neighborhoods (ablation) | H2GCN-1 · syn-products, heterophily | accuracy | Helped |
| Include 1-hop neighbors alongside the ego and 2-hop neighborhoods (ablation) | H2GCN-1 · syn-products | accuracy | Helped |
| Include 2-hop neighbors alongside the ego and 1-hop neighborhoods (ablation) | H2GCN-1 · syn-products | accuracy | Helped |
| Combine intermediate representations from all rounds (ablation) | H2GCN-2 · syn-products | accuracy | Helped |
| Add the first-round intermediate representation to the final representation (ablation) | H2GCN-2 · syn-products | accuracy | Helped |
| Add the second-round intermediate representation to the final representation (ablation) | H2GCN-2 · syn-products | accuracy | Helped |
| Combine the ego embedding with two-round intermediate representations (ablation) | GCN · syn-cora, h=0.1 | accuracy: 20% | Helped |
| Add jumping-knowledge intermediate representations (ablation) | GCN · heterophily benchmarks | accuracy: up to 14% | Helped |
| Add jumping-knowledge intermediate representations (ablation) | GCN-Cheby · heterophily benchmarks | accuracy: up to 8% | Helped |
| Add jumping-knowledge intermediate representations (ablation) | GraphSAGE · heterophily benchmarks | accuracy | No clear effect |
| Add jumping-knowledge intermediate representations (ablation) | GraphSAGE, GCN, and GCN-Cheby · homophily benchmarks | accuracy: gaps mostly less than 2% | No clear effect |
Closest papers by meaning
- Is Homophily a Necessity for Graph Neural Networks? (ICLR 2022 · similarity 0.95)
- On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks (NeurIPS 2024 · similarity 0.93)
- Understanding Heterophily for Graph Neural Networks (ICML 2024 · similarity 0.92)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.92)
- Integrating Co-Training with Edge Discrimination to Enhance Graph Neural Networks Under Heterophily (AAAI 2025 · similarity 0.92)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.92)
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.