Research map / Graph learning / GNN architectures
RADE: Random Add-Drop Edge as a Regularizer
Danial Saber, Amirali Salehi-Abari
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 |
|---|---|---|---|
| use RADE-OF instead of plain GCN | GCN · ogbn-arxiv | accuracy: 70.51 -> 71.09 | Improved |
| use RADE-OF instead of DropEdge | GCN · Cora | accuracy: 80.28 -> 81.08 | Improved |
| use RADE-OFS instead of GCN | GCN · peptides-func | AP: 55.14 -> 56.97 | Improved |
| use RADE-OFS instead of GCN | GCN · MUTAG | accuracy: 84.00 -> 86.24 | Improved |
| RADE-OFS vs RADE-OF on node classification | RADE · node classification (8 datasets) | accuracy | No clear effect |
| add GradNorm adaptive rates (ablation) | RADE-OF · Flickr | micro-F1: 51.42 -> 52.25 | Improved |
| add expectation-preserving correction (with fixed rates) (ablation) | RADE-OF · Flickr | micro-F1: 49.35 -> 51.42 | Improved |
| add expectation-preserving correction (with fixed rates) (ablation) | RADE-OF · Cora | accuracy: 78.94 -> 80.90 | Improved |
| use nonlinear GCN instead of linearized SGC backbone (ablation) | RADE-OF · Flickr | micro-F1: 50.65 -> 52.25 | Improved |
| combine add and drop instead of drop-only (ablation) | RADE-OF · ogbn-arxiv | accuracy: 70.68 -> 71.09 | Improved |
| combine add and drop instead of add-only (ablation) | RADE-OF · Cora | accuracy: 80.68 -> 81.08 | Improved |
| use adaptive GradNorm rates instead of tuned fixed rates (ablation) | RADE-OFS · Computer | accuracy: 88.31 -> 89.94 | Improved |
Closest papers by meaning
- Pruning Spurious Subgraphs for Graph Out-of-Distribution Generalization (NeurIPS 2025 · similarity 0.87)
- GDeR: Safeguarding Efficiency, Balancing, and Robustness via Prototypical Graph Pruning (NeurIPS 2024 · similarity 0.87)
- Effects of Dropout on Performance in Long-range Graph Learning Tasks (NeurIPS 2025 · similarity 0.86)
- Spectral Graph Pruning Against Over-Squashing and Over-Smoothing (NeurIPS 2024 · similarity 0.86)
- On the Rademacher Complexity of Graph Neural Networks: Unifying Expressivity and Geometry (NeurIPS 2026 · similarity 0.86)
- Graph Neural Network Generalization With Gaussian Mixture Model Based Augmentation (ICML 2025 · 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.
