Research map / Graph learning / Knowledge and temporal graphs
UrbanGraph: Physics-Informed Spatio-Temporal Dynamic Heterogeneous Graphs for Urban Microclimate Prediction
Weilin Xin, Chenyu Huang, Peilin Li, Jing Zhong, Jiawei Yao
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 heterogeneous relation-specific message passing (RGCN) instead of homogeneous GCN (ablation) | UrbanGraph · Ablation (hidden 384, single head) | R2: 0.8336 -> 0.8629 | Improved |
| use dynamic time-varying graph instead of static graph (ablation) | UrbanGraph · Ablation (hidden 384, single head) | R2: 0.8057 -> 0.8629 | Improved |
| inject physics as hard graph topology rather than PINN soft loss (ablation) | GCN+LSTM backbone · Physics injection strategy (GCN+LSTM backbone) | Avg R2: 0.7876 -> 0.8542 | Improved |
| add PDE soft-constraint loss (ablation) | GCN+LSTM · Physics injection strategy (GCN+LSTM backbone) | Avg R2: 0.7846 -> 0.7876 | No clear effect |
| explicit causal pruning vs implicit recurrent dynamic graph (LRGCN) | UrbanGraph vs LRGCN · UMC4/12 test set (avg over 6 variables), main comparison (hidden 128) | Avg R2: 0.8422 -> 0.8542 | Improved |
| explicit causal pruning vs LRGCN | UrbanGraph vs LRGCN · UMC4/12 test set (avg over 6 variables), main comparison (hidden 128) | FLOPs: 3.49e10 -> 9.13e9 | Improved |
| use concatenation fusion instead of attention or multiplicative (ablation) | UrbanGraph · Fusion strategy | Avg R2: 0.8491/0.8515 -> 0.8542 | Improved |
| add LSTM temporal module (ablation) | UrbanGraph · 12-hour horizon | R2 | Improved |
| use single prediction head instead of multi-head (ablation) | UrbanGraph · ablation setting | R2 | Improved |
| add UrbanGraph vs STGCN efficiency | UrbanGraph vs STGCN · UMC4/12 test set (avg over 6 variables), main comparison (hidden 128) | training time (s/epoch): 5.3547 -> 24.4823 | Hurt |
Closest papers by meaning
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.88)
- Causal Spatio-Temporal Prediction: An Effective and Efficient Multi-Modal Approach (NeurIPS 2025 · similarity 0.87)
- Knowledge Graph Guided Heterogeneity-Informed Diffusion Model for Spatio-Temporal Generation (AAAI 2026 · similarity 0.87)
- Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model (NeurIPS 2024 · similarity 0.86)
- ST-FiT: Inductive Spatial-Temporal Forecasting with Limited Training Data (AAAI 2025 · similarity 0.85)
- Diffusion Transformers as Open-World Spatiotemporal Foundation Models (NeurIPS 2025 · similarity 0.85)
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.
