Implicit Neural Representations via Spatially Indexed Time Series

Replacing dense spatiotemporal sampling with sequence-level supervision over spatially indexed time series reduces computational costs for time-varying volumetric data. This method learns the full temporal evolution of each spatial location instead of using coordinate-wise scalar samples.