Research map / Graph learning / GNN architectures
Discrete Structure Augmentation for Graph Convolutional Networks
Jianxin Ren, Weining Wu
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 graph reconstruction loss (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 73.10 without (Cora); 71.72 -> 64.25 (Citeseer) | Improved |
| add edge predictor (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 76.85 without (Cora); 71.72 -> 67.61 (Citeseer) | Improved |
| add unlabeled-to-labeled edge augmentation (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 74.62 without (Cora); 71.72 -> 65.90 (Citeseer) | Improved |
| use variational EM framework (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 74.95 without (Cora); 71.72 -> 52.57 (Citeseer) | Improved |
| add E-step modeling and optimization (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 77.56 without (Cora); 71.72 -> 67.60 (Citeseer) | Improved |
| iterative training instead of one-shot inference (ablation) | DSA-GCN · Cora/Citeseer (60% edges removed) | accuracy (%): 79.18 -> 76.30 one-shot (Cora); 71.72 -> 67.20 (Citeseer) | Improved |
| link unlabeled nodes to labeled nodes (VU-VL augmentation) (ablation) | DSA-GCN · Cora (20% edges removed) | accuracy (%): 78.2 -> 80.9 | Improved |
| link unlabeled nodes to similar unlabeled nodes (VU-VU augmentation) (ablation) | DSA-GCN · Cora (20% edges removed) | accuracy (%): 78.2 -> 79.4 | Improved |
| link unlabeled nodes to labeled nodes (VU-VL augmentation) (ablation) | DSA-GCN · CiteSeer (60% edges removed) | accuracy (%): 63.4 -> 68.2 | Improved |
| DSA-GCN vs GCN under extreme sparsity | DSA-GCN · Cora (extremely sparse, 60% edges dropped) | accuracy (%): 52.6 -> 75.4 | Improved |
| DSA-GCN vs best baseline under extreme sparsity | DSA-GCN · Amazon-Photo (extremely sparse) | accuracy (%): 93.1 (LSGNN) vs 92.8 | No clear effect |
| DSA-GCN vs best baseline on Pubmed under extreme sparsity | DSA-GCN · Pubmed (extremely sparse) | accuracy (%): 80.9 (MGCN) vs 78.4 | Hurt |
Closest papers by meaning
- Learning Discrete Structures for Graph Neural Networks (ICML 2019 · similarity 0.91)
- Bayesian Graph Convolutional Neural Networks for Semi-Supervised Classification (AAAI 2019 · similarity 0.88)
- Deeper Insights Into Graph Convolutional Networks for Semi-Supervised Learning (AAAI 2018 · similarity 0.87)
- Deep Multi-modal Graph Clustering via Graph Transformer Network (AAAI 2025 · similarity 0.86)
- Through the Dual-Prism: A Spectral Perspective on Graph Data Augmentation for Graph Classifications (AAAI 2025 · similarity 0.86)
- 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.
