Expand description
Recursive Feature Elimination (RFE) and RFE with Cross-Validation (RFECV).
RFE recursively removes the least-important features, ranking features
by their importance at each elimination step. The importance is determined by
an external importance vector that the user supplies via a callback.
RFECV extends RFE by using cross-validation to find the optimal number
of features to retain.
Because ferrolearn-preprocess cannot depend on estimator crates (to avoid
circular dependencies), these implementations accept feature importance
vectors directly rather than wrapping fitted estimators.