Research map / Graph learning / Graph self-supervised learning
Metropolis-Hastings Data Augmentation for Graph Neural Networks
Hyeonjin Park, Seunghun Lee, Sihyeon Kim, Jinyoung Park, Jisu Jeong, Kyung-Min Kim, Jung-Woo Ha, Hyunwoo J. Kim
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 MH-Aug augmentation with consistency and entropy regularizers (w/ Reg) | vanilla GCN/GraphSAGE/GAT · five node classification datasets | accuracy: 3.16% on average | Improved |
| MH-Aug (w/ Reg) instead of vanilla | GraphSAGE · CORA | accuracy: 79.78 -> 84.70 | Improved |
| MH-Aug (w/ Reg) instead of vanilla | GCN, GraphSAGE, GAT · CITESEER | accuracy: 4.16% | Improved |
| remove regularizers (supervised only, MH-Aug w/o Reg) vs vanilla (ablation) | MH-Aug · five datasets | accuracy: 1.47% on average | Improved |
| MH-Aug (w/o Reg) instead of vanilla GAT (ablation) | GAT · Computers | accuracy: 78.47 -> 81.72 | Improved |
| MH-Aug (w/o Reg) instead of DropEdge | GCN · CORA | accuracy: 82.21 -> 83.55 | Improved |
| add consistency and entropy regularizers to MH-Aug (ablation) | MH-Aug with GCN · CORA | accuracy: 83.55 -> 85.16 | Improved |
| add regularizers to MH-Aug (ablation) | MH-Aug with GraphSAGE · Photo | accuracy: 90.37 -> 92.19 | Improved |
| AdaEdge instead of vanilla | GAT · CORA | accuracy: 82.23 -> 77.90 | Hurt |
| remove normalization term by number of possible augmented graphs from target distribution (ablation) | MH-Aug target distribution · fully connected 100-node graph | mean ego-graph change ratio | Hurt |
| increase sigma_E to 1000 (ego-graph diversity) (ablation) | MH-Aug · CORA ego-graph samples | edge-drop probability | Mixed |
Closest papers by meaning
- Regularizing Graph Neural Networks via Consistency-Diversity Graph Augmentations (AAAI 2022 · similarity 0.91)
- Local Augmentation for Graph Neural Networks (ICML 2022 · similarity 0.90)
- Graph Neural Network Generalization With Gaussian Mixture Model Based Augmentation (ICML 2025 · similarity 0.88)
- Through the Dual-Prism: A Spectral Perspective on Graph Data Augmentation for Graph Classifications (AAAI 2025 · similarity 0.87)
- Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification (AAAI 2025 · similarity 0.86)
- Community-Invariant Graph Contrastive Learning (ICML 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.
