RFE
Recursive Feature Elimination: repeatedly fit an estimator and discard the weakest feature until the target count remains.
Algorithm
At each round the base estimator is refit on the surviving features and the one with the smallest absolute weight/importance is removed, yielding a ranking of all features.
Constructor
Skigen::RFE<Estimator> model(Estimator est, int n_features_to_select);
Parameters
| Parameter | Default | Description |
|---|---|---|
estimator | — | Estimator exposing weights/importances. |
n_features_to_select | — | Number of features to keep. |
Methods
| Method | Description |
|---|---|
fit(X, y) | Run the elimination loop. |
transform(X) | Project onto the selected features. |
get_support_mask() | Boolean mask of selected features. |
Fitted Attributes
| Accessor | Description |
|---|---|
ranking() | Elimination rank of each feature (1 = kept). |
Example
Skigen::RFE rfe(Skigen::Ridge<double>(0.1), 3);
rfe.fit(X, y);
auto X_sel = rfe.transform(X);
Verified by unit tests
Covered by the module's CTest suite under tests/.
API Reference
For full signatures see the RFE API Reference.