Research map / Graph learning / Graph self-supervised learning
Disentangled Graph Convolutional 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 |
|---|---|---|---|
| Replace holistic GCN-style aggregation with DisenConv neighborhood routing (disentangled channels) | DisenGCN · Semi-supervised node classification (Cora, Citeseer, Pubmed) | accuracy (%): +2.2 Cora, +3.1 Citeseer, +1.5 Pubmed over GCN | Helped |
| Disentangled routing model with 5 layers instead of 2-layer baselines | DisenGCN · Semi-supervised node classification | accuracy (%): outperforms GAT on all three datasets | Helped |
| Disentangled routing instead of GCN on synthetic graphs with many latent factors | DisenGCN · Synthetic graphs, 4 to 16 latent factors | Micro-F1 (%): +12.02% to +17.63% relative improvement over GCN across K=4..16 | Helped |
| Increase number of latent factors in synthetic graph (4 to 10 vs. K>12) | DisenGCN relative to GCN · Synthetic graphs | Micro-F1 relative improvement: gain grows from +12.02% (K=4) to +17.63% (K=10), then falls to +12.6% (K=16) | Mixed |
| Number of channels set near the true number of latent factors (vs. far from it) (ablation) | DisenGCN · Synthetic graph with eight latent factors | Micro-F1 (%) | Mixed |
| More neighborhood routing iterations (ablation) | DisenGCN · Synthetic graph with eight latent factors | Micro-F1 (%) | Helped |
| Multi-label node classification (BlogCatalog, PPI, POS) versus GCN/GAT/node embeddings | DisenGCN · Multi-label node classification | Macro-F1 and Micro-F1: around 10% to 20% relative improvement in most cases | Helped |
Closest papers by meaning
- Independence Promoted Graph Disentangled Networks (AAAI 2020 · similarity 0.95)
- Factorizable Graph Convolutional Networks (NeurIPS 2020 · similarity 0.92)
- Disentangled Graph Self-supervised Learning for Out-of-Distribution Generalization (ICML 2024 · similarity 0.88)
- Enhancing Size Generalization in Graph Neural Networks through Disentangled Representation Learning (ICML 2024 · similarity 0.88)
- Disentangling, Amplifying, and Debiasing: Learning Disentangled Representations for Fair Graph Neural Networks (AAAI 2025 · similarity 0.87)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · 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.