PatchSTG partitions sensors into locality-preserving geographic patches and applies dual intra/inter-patch attention to reduce spatiotemporal modeling complexity from quadratic to near-linear while maintaining competitive traffic forecast accuracy.
Yuan and Li [19] identify persistent gaps—limited real-time scalability, high computational cost, and difficulty modeling heterogeneous spatial structures
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
PatchSTG: Scalable Spatiotemporal Graph Transformers for Traffic Forecasting on Irregular Sensor Networks
PatchSTG partitions sensors into locality-preserving geographic patches and applies dual intra/inter-patch attention to reduce spatiotemporal modeling complexity from quadratic to near-linear while maintaining competitive traffic forecast accuracy.