Research map / Graph learning / GNN architectures
Graph Neural Networks Inspired by Classical Iterative Algorithms
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 |
|---|---|---|---|
| Add IRLS-based attention to TWIRLS base | TWIRLS · ATK-Cora | accuracy: 62.89 ± 1.59 -> 70.23 ± 1.09 | Helped |
| Add IRLS-based attention to TWIRLS base | TWIRLS · ATK-Citeseer | accuracy: 63.83 ± 1.95 -> 70.63 ± 0.93 | Helped |
| Add IRLS-based attention to TWIRLS base | TWIRLS · Texas, Wisconsin, Actor, Cornell | accuracy | Mixed |
| Add IRLS-based attention to TWIRLS base | TWIRLS · Texas | accuracy: 81.62±5.51 -> 84.59±3.83 | Helped |
| Add IRLS-based attention to TWIRLS base | TWIRLS · Wisconsin | accuracy: 82.75±7.83 -> 86.67±4.19 | Helped |
| Add IRLS-based attention to TWIRLS base | TWIRLS · Actor | accuracy: 37.10±1.07 -> 37.43±1.50 | Helped |
| Add IRLS-based attention to TWIRLS base | TWIRLS · Cornell | accuracy: 83.51±7.30 -> 86.76±5.05 | Helped |
| Match propagation step count to gradient step size (ablation) | TWIRLS base · Citeseer | accuracy: 66.00 -> 74.07 | Helped |
| Set attention truncation T=2 rather than T=+∞ (ablation) | TWIRLS · ATK-Cora | accuracy: 69.77 ± 1.26 -> 70.23 ± 1.09 | Helped |
| Set attention truncation T=2 rather than T=+∞ (ablation) | TWIRLS · ATK-Citeseer | accuracy: 70.51 ± 1.09 -> 70.63 ± 0.93 | No clear effect |
| Use 2 rather than 1 MLP layer before propagation (ablation) | TWIRLS base · Cora | accuracy: 83.3 ± 0.3 -> 84.1 ± 0.5 | Helped |
| Use 2 rather than 1 MLP layer before propagation (ablation) | TWIRLS base · Citeseer | accuracy: 74.1 ± 0.5 -> 74.2 ± 0.45 | No clear effect |
Closest papers by meaning
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.92)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.91)
- On Vanishing Gradients, Over-Smoothing, and Over-Squashing in GNNs: Bridging Recurrent and Graph Learning (NeurIPS 2025 · similarity 0.91)
- On the Stability of Graph Convolutional Neural Networks: A Probabilistic Perspective (NeurIPS 2025 · similarity 0.91)
- Classic GNNs are Strong Baselines: Reassessing GNNs for Node Classification (NeurIPS 2024 · similarity 0.90)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 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.