Basic classes to combine transformers and estimators

[agoscinski]

Implementation of the TransformedTargetRegressor and Pipeline. Their usage is shown in the composed-model.py (basically a copy of the linear model without weighting). I get as error values

RMSE energies = 121.938 eV
RMSE forces = 0.094 eV/Å

in contrast to the Ridge alone

RMSE energies = 0.163 eV
RMSE forces = 0.267 eV/Å

Seems not totally unreasonable considering that I used the same alpha values. Started to write test, but I did not have time to finish them. I got an error which I did get stuck on.

Missing:

• !tests!
• passing on arguments like sample_weights
• base classes for transformers and estimators to implement checks

Side note: The sklearn Pipeline works for fitting and predicting out of the box. The score function only does not work.

[PicoCentauri]

Close because we found a better solution in #10