You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you come up with a nice way to flatten (and then reshape back) your matrix/image into TimeSeries components, it should be possible? However, due to considerable dimensions increase, the training will definitely take much more time. Using input_chunk_length>1 will allow you to intake several images in the past, and output_chunk_length>1 to forecast several images into the future. The covariates could probably still be used.
I don't think that any of the model in Darts was implemented with such use-case in mind but I am very curious to see if you can make it work.
Is it possible to use the FFT forecasting model with matrices, for example forecasting image continuation?
The text was updated successfully, but these errors were encountered: