Tag
This paper proposes a framework that uses entropy-based diagnostics to harmonize spatial and temporal feature representations, achieving substantial accuracy gains on large-scale spatiotemporal prediction tasks across urban traffic, meteorology, and epidemic datasets.