How to handle windowing ?

[raullalves]

First of all, great work !!

I'd like to know, how to proceed with windowing in your ESN approach ?

Maybe, adding one more weight layer at the beginning ?

Regards

[zimmerrol]

Hi, can you give an example of what you mean with windowing?

[raullalves]

Sorry about the delay, @zimmerrol

Regarding the PredictionESN, your approach works perfectly with series in the shape (x, n_input) and "labels" with shape (x, n_output)

However, for time series forecasting, to have the possibility of using validation for improving the results, it is common to divide the series in small windows with a fixed length.

For instance, if window size = 12, the input series will have shape (x', 12, n_input) and the labels with shape (x', n_output).

This approach is just like dividing the data in batches.

Do you think that the "windowing" approach is not necessary?

Regards