Research map / Graph learning / Graph self-supervised learning
Graph U-Nets
Hongyang Gao, Shuiwang Ji
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add gPool and gUnpool encoder-decoder to GCN | GCN · Cora | accuracy: 81.5% -> 84.4 ± 0.6% | Improved |
| add gPool and gUnpool encoder-decoder to GCN | GCN · Citeseer | accuracy: 70.3% -> 73.2 ± 0.5% | Improved |
| add gPool and gUnpool encoder-decoder to GCN | GCN · Pubmed | accuracy: 79.0% -> 79.6 ± 0.2% | Improved |
| add gPool and gUnpool layers (ablation) | g-U-Nets · Cora | accuracy: 82.1 ± 0.6% -> 84.4 ± 0.6% | Improved |
| add gPool and gUnpool layers (ablation) | g-U-Nets · Citeseer | accuracy: 71.6 ± 0.5% -> 73.2 ± 0.5% | Improved |
| add gPool and gUnpool layers (ablation) | g-U-Nets · Pubmed | accuracy: 79.1 ± 0.2% -> 79.6 ± 0.2% | Improved |
| add graph connectivity augmentation via 2nd graph power (ablation) | g-U-Nets · Cora | accuracy: 83.7 ± 0.7% -> 84.4 ± 0.6% | Improved |
| add graph connectivity augmentation via 2nd graph power (ablation) | g-U-Nets · Citeseer | accuracy: 72.5 ± 0.6% -> 73.2 ± 0.5% | Improved |
| add graph connectivity augmentation via 2nd graph power (ablation) | g-U-Nets · Pubmed | accuracy: 79.0 ± 0.3% -> 79.6 ± 0.2% | Improved |
| increase depth from 2 to 4 blocks (ablation) | g-U-Nets · Cora | accuracy: 82.6 ± 0.6% -> 84.4 ± 0.6% | Improved |
| increase depth from 4 to 5 blocks (ablation) | g-U-Nets · Cora | accuracy: 84.4 ± 0.6% -> 84.1 ± 0.5% | Hurt |
| increase depth from 4 to 5 blocks (ablation) | g-U-Nets · Citeseer | accuracy: 73.2 ± 0.5% -> 72.8 ± 0.6% | Hurt |
Closest papers by meaning
- Rethinking pooling in graph neural networks (NeurIPS 2020 · similarity 0.90)
- Beyond Convolution: Advancing Hypergraph Neural Networks with Hypergraph U-Nets (TPAMI 2026 · similarity 0.89)
- Geometry-Aware Edge Pooling for Graph Neural Networks (NeurIPS 2025 · similarity 0.89)
- SSHPool: The Separated Subgraph-based Hierarchical Pooling (AAAI 2026 · similarity 0.88)
- AttPool: Towards Hierarchical Feature Representation in Graph Convolutional Networks via Attention Mechanism (ICCV 2019 · similarity 0.88)
- Universal Graph Convolutional Networks (NeurIPS 2021 · similarity 0.88)
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.
