SelectFromModel
A meta-transformer that selects features by thresholding the coef_ or feature_importances_ of a wrapped estimator.
Algorithm
Fits the base estimator, then keeps features whose importance exceeds a threshold (mean, median, or a numeric value).
Constructor
Skigen::SelectFromModel<Estimator> model(Estimator est, Threshold = Mean);
Parameters
| Parameter | Default | Description |
|---|---|---|
estimator | — | Estimator exposing coefficients/importances. |
threshold | Mean | Mean, Median, or a numeric cutoff. |
Methods
| Method | Description |
|---|---|
fit(X, y) | Fit the base estimator and compute the mask. |
transform(X) | Keep features above the threshold. |
get_support_mask() | Boolean mask of selected features. |
Fitted Attributes
| Accessor | Description |
|---|---|
estimator() | The fitted base estimator. |
Example
Skigen::SelectFromModel sel(Skigen::Ridge<double>(0.1));
sel.fit(X, y);
auto X_sel = sel.transform(X);
Verified by unit tests
Covered by the module's CTest suite under tests/.
API Reference
For full signatures see the SelectFromModel API Reference.