Change model scoring to be more similar to scikit-learn and dask-searchcv

ContinuumIO / elm

Phase I & part of Phase II of NASA SBIR - Parallel Machine Learning on Satellite Data

http://ensemble-learning-models.readthedocs.io

44 stars 23 forks source link

Change model scoring to be more similar to scikit-learn and dask-searchcv #99

Closed PeterDSteinberg closed 7 years ago

PeterDSteinberg commented 8 years ago

[ ] Change the interface to fit_ensemble and fit_ea to include most or all of the arg spec of GridSearchCV:

dask_searchcv.GridSearchCV(estimator, param_grid, scoring=None, iid=True, refit=True, cv=None, error_score='raise', return_train_score=True, scheduler=None, n_jobs=-1, cache_cv=True)

[ ] Also change the reporting of model scores - see cv_results_ attribute here

PeterDSteinberg commented 7 years ago

See also dask-glm as part of our efforts of unifying our approaches to ML.

PeterDSteinberg commented 7 years ago

When we refactor elm to merge in the capabilities of dask-searchcv and dask-glm, then after that work is solidified we should address #171 and #182 ( cc @gbrener )

PeterDSteinberg commented 7 years ago

Fixed by #192 - Now EaSearchCV instances have a cv_results_ attribute like sklearn.model_selection.GridSearchCV