Search and cross-fitting
Tools for selecting a representation and for producing leakage-free supervised features. See Target awareness for the leakage model.
Representation search
Cross-validate a downstream estimator over candidate numerical methods and refit the best one.
Select the best numerical representation by cross-validation. |
Cross-fitting
Produce out-of-fold training features from a supervised transformer while transforming new data with an all-data model.
Cross-fit a supervised transformer to remove target leakage on training data. |