Time Series Model

Transformer Model forecasting

• Segmentation of time series into subseries-level patches which are served as input tokens to Transformer

• Channel-independence where each channel contains a single univariate time series that shares the same embedding and Transformer weights across all the series