Research map / Label-efficient and robust learning / Recognition and long-tail
Boosting Fine-Grained Urban Flow Inference via Lightweight Architecture and Focalized Optimization
Yuanshao Zhu, Xiangyu Zhao, Zijian Zhang, Xuetao Wei, James Jianqiao Yu
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 |
|---|---|---|---|
| add FiLM modulation (ablation) | PLGF · TaxiBJ Task-1 | MSE: 14.481 -> 14.408 | Improved |
| add Local-Global Fusion (ablation) | PLGF · TaxiBJ Task-1 | MSE: 17.745 -> 14.408 | Improved |
| add Context-Gated Attention (ablation) | PLGF · TaxiBJ Task-1 | MSE: 18.638 -> 14.408 | Improved |
| add DualFocal loss (ablation) | PLGF · TaxiBJ Task-1 | MAPE: 0.304 -> 0.281 | Improved |
| add DualFocal loss (ablation) | PLGF · TaxiBJ Task-1 | MSE: 14.574 -> 14.408 | Improved |
| add logarithmic term (ablation) | PLGF · TaxiBJ Task-1 | MAPE: 0.301 -> 0.281 | Improved |
| add logarithmic term (ablation) | PLGF · TaxiBJ Task-1 | MSE: 14.349 -> 14.408 | Hurt |
| plug DualFocal loss into UrbanFM | UrbanFM · TaxiBJ Task-2 | MSE: 4.4% | Improved |
| plug DualFocal loss into UrbanFM | UrbanFM · TaxiBJ Task-2 | MAPE: 8.3% | Improved |
| plug DualFocal loss into CUFAR | CUFAR · TaxiBJ Task-2 | MAPE: 8.0% | Improved |
| plug DualFocal loss into UNO | UNO · TaxiBJ Task-2 | MAPE: 7.4% | Improved |
| increase channels from 64 to 128 | PLGF · TaxiBJ | MSE | Improved |
Closest papers by meaning
- Model-Informed Flows for Bayesian Inference (NeurIPS 2025 · similarity 0.82)
- Drawing Informative Gradients from Sources: A One-stage Transfer Learning Framework for Cross-city Spatiotemporal Forecasting (AAAI 2025 · similarity 0.82)
- BO4Mob: Bayesian Optimization Benchmarks for High-Dimensional Urban Mobility Problem (NeurIPS 2025 · similarity 0.82)
- Closed-Loop Supervised Fine-Tuning of Tokenized Traffic Models (CVPR 2025 · similarity 0.82)
- Spatio-Temporal Few-Shot Learning via Diffusive Neural Network Generation (ICLR 2024 · similarity 0.82)
- Inference-Time Scaling for Flow Models via Stochastic Generation and Rollover Budget Forcing (NeurIPS 2025 · similarity 0.82)
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.
