I have a multiple datasets that describe the status of independent engines and I would like to create a model that identifies anomalies over the given time-series like a LSTM-VAE or USAD, the point is that I do not know how to make the model to generalize over multiple independent datasets, ie how to create a well formatted Dataloader.
One of the possible idea that had come in my mind is set the batch to be equal of the number of engines and then iterate over the batch splitted by the sequence length, since it is unfeasible to stack all the datasets into a single one since the model window possibly get different engines timestamp.
Thanks for your help!
Best regards
Riccardo