Research map / Graph learning / GNN architectures
Defending Graph Convolutional Networks against Dynamic Graph Perturbations via Bayesian Self-Supervision
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 |
|---|---|---|---|
| Apply GraphSS label-transition recovery after the L&F attack | GCN · Cora, Citeseer, Pubmed, AMZcobuy, Coauthor | test accuracy (%): 17.01 (±5.11) -> 65.38 (±2.59); 35.42 (±1.67) -> 69.18 (±0.99); 41.49 (±1.99) -> 81.16 (±1.41); 46.71 (±1.95) -> 90.88 ( | Helped |
| Apply GraphSS label-transition recovery after the L&F attack | SGC · Cora, Citeseer, Pubmed, AMZcobuy, Coauthor | test accuracy (%): 36.55 (±8.16) -> 88.51 (±2.68); 44.79 (±2.75) -> 70.78 (±3.17); 42.60 (±1.95) -> 86.06 (±1.29); 45.34 (±2.02) -> 88.70 ( | Helped |
| Apply GraphSS label-transition recovery after the L&F attack | GraphSAGE · Cora, Citeseer, Pubmed, AMZcobuy, Coauthor | test accuracy (%): 36.48 (±7.72) -> 89.29 (±2.77); 42.09 (±3.94) -> 73.43 (±2.32); 44.63 (±2.19) -> 87.55 (±1.16); 45.87 (±2.20) -> 90.20 ( | Helped |
| Remove dynamic updates to the label-transition matrix φ (ablation) | GraphSS on GCN · Five graph datasets under L&F perturbations | test accuracy | No clear effect |
| Add classifier retraining at each inference iteration, alongside dynamic φ (ablation) | GraphSS on GCN · Five graph datasets under L&F perturbations | test accuracy | Helped |
| Vary warm-up steps WS away from the selected optimum (ablation) | GraphSS with GCN · Validation set | validation accuracy | Hurt |
| Vary retraining epochs away from the selected optimum (ablation) | GraphSS with GCN · Validation set | validation accuracy | Hurt |
Closest papers by meaning
- When Does Self-Supervision Help Graph Convolutional Networks? (ICML 2020 · similarity 0.91)
- Bayesian Graph Convolutional Neural Networks for Semi-Supervised Classification (AAAI 2019 · similarity 0.90)
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective (NeurIPS 2025 · similarity 0.90)
- Deeper Insights Into Graph Convolutional Networks for Semi-Supervised Learning (AAAI 2018 · similarity 0.89)
- L2-GCN: Layer-Wise and Learned Efficient Training of Graph Convolutional Networks (CVPR 2020 · similarity 0.89)
- On Which Nodes Does GCN Fail? Enhancing GCN From the Node Perspective (ICML 2024 · 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.