Add `Reconstruction` method for spatio-temporal embedding about dynamicalsystemsbase.jl HOT 2 CLOSED

I am not so sure about whether this should be implemented as a feature of the package. It is really pushing stuff in the limit of doing everything for the user and the user doing nothing.

It is already possible (and in my eyes quite easy) to do what you are reporting:

1. The user can successively choose the columns of this matrix and reconstruct them individually. Let's say he chooses columns 4,5,6 corresponding to I = 1, K = 1. Another example would be choosing N-1, N, c. Then reconstruct this dataset with D, τ.

2. Then predict this dataset. Then, choose the midle column of the prediction as the timeseries to keep as a prediction. This is the prediction of the spatiotemporal column 5.

3. Repeat the process for all columns.

Can you show whether there is any gain in implementing what you suggest as a feature of the package?

(yes I know that you have to do it irrespectively for your thesis)

from dynamicalsystemsbase.jl.

I think we have to be careful so that the packages make it easy for users to do science, but don't do the science for them.

This can be risky.

from dynamicalsystemsbase.jl.