Spatiotemporal Forecasting

This video animates a neural network architecture for time-series forecasting, walking through data normalization, gated temporal convolutions, dynamic spatial attention, and multi-scale aggregation. It concludes by showing how the model generates future predictions from historical input data.