Map different preprocess to estimators about mlens HOT 3 CLOSED

@shuyangdu no that's not what happens. Think of the list passed in the preprocessing_cases dict as one transformer (essentially a Scikit-learn pipeline), and all the estimators in the list under the same key in the estimators dict as all estimators applied to the transformed data. Essentially like this:

X_tr = X.copy()
for transformer in preprocessing_cases['case-1']:
    X_tr = trans_1.fit_transform(X_tr)

for estimator in estimators['case-1']:
    estimator.fit(X_tr, y)

Note that if you only have one preprocessing case, you don't need to create a dictionary but can pass the lists directly.

Sorry for the confusion, hope this clarifies.

from mlens.

Thank you for your reply! So to verify, the list for preprocessing and estimators have different meanings. The list for peprocessing is for 'sequentially' preprocess data. But the list for estimators is just a set of estimators. The order in preprocessing list matters while the order in estimator list not.

from mlens.

Happy to help, and yes that's right!

from mlens.