Research map / Graph learning / GNN architectures
Is Homophily a Necessity 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 GCN instead of MLP | Semi-supervised node classification · Chameleon | accuracy: 48.11 ± 2.23 -> 67.96 ± 1.82 | Helped |
| Use GCN instead of MLP | Semi-supervised node classification · Squirrel | accuracy: 31.68 ± 1.90 -> 54.47 ± 1.17 | Helped |
| Use GCN instead of MLP | Semi-supervised node classification · Cora | accuracy: 75.04 ± 1.97 -> 87.12 ± 1.38 | Helped |
| Use GCN instead of MLP | Semi-supervised node classification · Citeseer | accuracy: 72.40 ± 1.97 -> 76.50 ± 1.61 | Helped |
| Use GCN instead of MLP | Semi-supervised node classification · Pubmed | accuracy: 87.84 ± 0.30 -> 88.52 ± 0.41 | Helped |
| Use GCN instead of MLP | Semi-supervised node classification · Actor | accuracy: 36.17 ± 1.09 -> 30.31 ± 0.98 | Hurt |
| Use GCN instead of MLP | Semi-supervised node classification · Cornell | accuracy: 84.86 ± 6.04 -> 59.35 ± 4.19 | Hurt |
| Use GCN instead of MLP | Semi-supervised node classification · Wisconsin | accuracy: 86.29 ± 4.50 -> 61.76 ± 6.15 | Hurt |
| Use GCN instead of MLP | Semi-supervised node classification · Texas | accuracy: 83.30 ± 4.54 -> 63.81 ± 5.27 | Hurt |
| Add targeted cross-label edges following distinct class-specific neighborhood distributions | GCN on Cora · Cora synthetic graph, h=0.25 | accuracy: 86% | Helped |
| Add targeted cross-label edges, increasing their number K | GCN on synthetic graphs · Cora and Citeseer synthetic graphs | accuracy | Mixed |
| Increase random-noise probability gamma in added edges, making class neighborhood distributions more similar | GCN on synthetic graphs · Cora synthetic graph at h=0.25 | accuracy | Hurt |
Closest papers by meaning
- Understanding Heterophily for Graph Neural Networks (ICML 2024 · similarity 0.93)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.93)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.93)
- On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks (NeurIPS 2024 · similarity 0.92)
- Refining Latent Homophilic Structures over Heterophilic Graphs for Robust Graph Convolution Networks (AAAI 2024 · similarity 0.92)
- Integrating Co-Training with Edge Discrimination to Enhance Graph Neural Networks Under Heterophily (AAAI 2025 · similarity 0.91)
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.