please help me ,what is the difference between windows and recurrent or multivariate in sampling?

Hi @ql h! They are used for different types of models: 1. windows: sample random windows of size input + h, starting at different timestamps. Used for univariate models with fixed input size such as mlp, nbeats, and transformers. 2. Recurrent: sample consecutive windows for each time series, used for models that predict windows sequentially like lstms. 3. Multivariate: sample random timestamps and include the window of size input + h starting at those timestamps, including all time series in the dataset. Used for multivariate models.