Passing fit_params to evaluations

[nguigs]

trafficstars

Hi!

Thank you for this great framework!

Would it be possible to pass a fit_params argument to the evaluate method, similarly to cross_val_score in scikit-learn? This would allow to propagate arguments to the fit methods of the pipeline/estimator being evaluated.

[sylvchev]

Yes, this is somethng we are considering to support transfer learning (see #119). I already work on some code outside MOABB to do that, it is quite easy. We could work on a PR to include it in MOABB if you want.

I think the best approach is to create a new evaluation for transfer learning, based on learning curves. I prefer not to expose the fit_params in general as it is quite specific for transfer learning.