Scalable Pre-Training of Compact Urban Spatio-Temporal Predictive Models on Large-Scale Multi-Domain Data
Summary: CompactST: a 300K-parameter pre-trained STP model trained on 300M spatio-temporal points across 10+ domains to enable robust few/zero-shot urban prediction in data-scarce settings. Novel mixture-of-normalizers, multi-scale spatio-temporal mixer, and dataset-oriented adaptive tuning handle domain/spatial heterogeneity and resolution mismatch, shifting dataset-specific parameters to fine-tuning for compact, efficient transfer. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Jindong Han (Shandong University)
- 2. Hao Wang (Hong Kong University of Science and Technology)
- 3. Hui Xiong (Hong Kong University of Science and Technology)
- 4. Hao Liu (Hong Kong University of Science and Technology)
BibTeX Citation
@article{han_vldb25,
title = {{Scalable Pre-Training of Compact Urban Spatio-Temporal Predictive Models on Large-Scale Multi-Domain Data}},
author = {Han, Jindong and Wang, Hao and Xiong, Hui and Liu, Hao},
journal = {PVLDB},
series = {{VLDB} '25},
volume = {18},
number = {7},
pages = {2149--2158},
doi = {10.14778/3734839.3734851},
url = {https://doi.org/10.14778/3734839.3734851},
year = {2025}
}Incoming Citations (Sorted by Pagerank)
Showing 0 of 0 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|
Outgoing Citations (Sorted by Pagerank)
Showing 9 of 9 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.