Research map / Graph learning / GNN architectures
Random Walk Conformer: Learning Graph Representation from Long and Short Range
Pei-Kai Yeh, Hsi-Wen Chen, Ming-Syan Chen
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 |
|---|---|---|---|
| Combine random-walk self-attention and random-walk convolution | RWC · ZINC | MAE: 0.0790 -> 0.0678 | Improved |
| Combine random-walk self-attention and random-walk convolution | RWC · PTC | accuracy: 68.2 ± 7.2 -> 74.4 ± 4.4 | Improved |
| Combine random-walk self-attention and random-walk convolution | RWC · NCI1 | accuracy: 86.0 ± 1.8 -> 82.2 ± 1.5 | Mixed |
| Add random-walk convolution to weighted RW-SAN with random-walk encoding (ablation) | RWC · ZINC | MAE: 0.116 -> 0.068 | Improved |
| Add random-walk convolution to weighted RW-SAN with random-walk encoding (ablation) | RWC · NCI1 | accuracy: 75.8 -> 82.2 | Improved |
| Use random-walk encoding instead of shortest-path encoding (ablation) | RWC · ZINC | MAE: 0.082 -> 0.068 | Improved |
| Use random-walk encoding instead of shortest-path encoding (ablation) | RWC · NCI1 | accuracy: 82.4 -> 82.2 | Mixed |
| Weight RW-SAN attention by node occurrence in the walk (ablation) | RWC · ZINC | MAE: 0.070 -> 0.068 | Improved |
| Weight RW-SAN attention by node occurrence in the walk (ablation) | RWC · NCI1 | accuracy: 82.6 -> 82.2 | Hurt |
| Use non-backtracking instead of backtracking random walks (ablation) | RWC · ZINC | MAE: 0.299 -> 0.191; 0.131 -> 0.068; 0.091 -> 0.065; 0.085 -> 0.077 | Improved |
| Increase training walk length from 20 to 50 (ablation) | RWC · ZINC | MAE: 0.191 -> 0.068 | Improved |
| Increase training walk length from 100 to 200 (ablation) | RWC · ZINC | MAE: 0.065 -> 0.077 | Hurt |
Closest papers by meaning
- Revisiting Random Walks for Learning on Graphs (ICLR 2025 · similarity 0.91)
- Random Search Neural Networks for Efficient and Expressive Graph Learning (NeurIPS 2025 · similarity 0.90)
- Learning Long Range Dependencies on Graphs via Random Walks (ICLR 2025 · similarity 0.89)
- Non-convolutional graph neural networks. (NeurIPS 2024 · similarity 0.88)
- Similarity-Navigated Conformal Prediction for Graph Neural Networks (NeurIPS 2024 · similarity 0.87)
- Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality (ICML 2025 · similarity 0.87)
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.
