Research map / Graph learning / GNN architectures
Graph Random Neural Networks for Semi-Supervised Learning on Graphs
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 |
|---|---|---|---|
| Use GRAND instead of GCN | Semi-supervised node classification · Cora-Full, Coauthor CS, Coauthor Physics, Amazon Computers, Amazon Photo, Aminer CS | accuracy: 62.2 $\pm$ 0.6 -> 63.5 $\pm$0.6; 91.1 $\pm$ 0.5 -> 92.9 $\pm$ 0.5; 92.8 $\pm$ 1.0 -> 94.6 $\pm$ 0.5; 82.6 $\pm$ 2.4 -> 8 | Helped |
| Use GRAND instead of the strongest listed baseline | Semi-supervised node classification · Cora at 1%, 3%, and 5% label rates; Citeseer at 1%, 3%, and 5%; Pubmed at 0.1%, 0.3%, and 0.5% | accuracy: 64.3$\pm$5.8 -> 69.1$\pm$4.0; 77.2$\pm$2.4 -> 79.5$\pm$2.2; 80.8$\pm$2.1 -> 83.0$\pm$1.6; 64.4$\pm$2.9 -> 65.3$\pm$3.3; | Helped |
| Use DropNode rather than feature-wise dropout for random propagation (ablation) | GRAND · Cora, Citeseer, Pubmed | accuracy | Helped |
| Use an MLP classifier rather than GCN or GAT in the GRAND framework (ablation) | GRAND · Graph node classification | accuracy | Helped |
| Add consistency regularization to random propagation (ablation) | GRAND · Semi-supervised node classification | generalization | Helped |
| Use mixed-order propagation instead of only the highest-order propagation (ablation) | GRAND · Graph node classification at larger propagation steps | accuracy | Helped |
| Add GRAND's random propagation and consistency regularization rather than removing both (ablation) | GRAND · Training and validation loss | cross-entropy loss gap | Helped |
| Increase the propagation step K (ablation) | GRAND · Cora | accuracy | Helped |
| Increase augmentation count S (ablation) | GRAND · Cora | accuracy | Helped |
| Increase the consistency-regularization coefficient λ (ablation) | GRAND and its variants · Cora | accuracy | Helped |
| Set DropNode probability δ to 0.5 rather than other tested values (ablation) | GRAND · Cora | accuracy | Helped |
| Increase propagation depth in GCN or GAT instead of using GRAND's separated propagation and MLP (ablation) | GCN and GAT · Cora | accuracy: 0.75 to 0.2 | Hurt |
Closest papers by meaning
- Regularizing Graph Neural Networks via Consistency-Diversity Graph Augmentations (AAAI 2022 · similarity 0.93)
- Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training data (NeurIPS 2021 · similarity 0.91)
- What functions can Graph Neural Networks compute on random graphs? The role of Positional Encoding (NeurIPS 2023 · similarity 0.91)
- Almost Surely Asymptotically Constant Graph Neural Networks (NeurIPS 2024 · similarity 0.90)
- Towards Bridging Generalization and Expressivity of Graph Neural Networks (ICLR 2025 · similarity 0.90)
- Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification (AAAI 2025 · similarity 0.90)
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.