SCAD<T>

Models & Types ยท AiDotNet.Preprocessing.FeatureSelection.Embedded

Smoothly Clipped Absolute Deviation (SCAD) penalty for feature selection.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> โ†’ SCAD<T>

Implements: IFixedCountFeatureSelector<T>, ISupervisedFeatureSelector<T>

For Beginners

LASSO tends to shrink large coefficients too much (bias). SCAD fixes this by penalizing large coefficients less aggressively while still shrinking small ones to zero. This gives you better estimates of important feature effects.

How It Works

SCAD is a non-convex penalty that addresses the bias problem of LASSO. It applies the same penalty as LASSO for small coefficients but reduces the penalty for large coefficients, leading to nearly unbiased estimates.

Constructors

Constructor Summary
SCAD(int nFeaturesToSelect, double lambda, double a, int maxIterations, double tolerance, int[] columnIndices)

Properties

Property Summary
Vector<T> Coefficients { get; }
bool SupportsInverseTransform { get; }
double A { get; }
double Lambda { get; }
int NFeaturesToSelect { get; }
int[] SelectedIndices { get; }

Methods

Method Summary
Matrix<T> FitTransform(Matrix<T> data, Vector<T> target)
bool[] GetSupportMask()
string[] GetFeatureNamesOut(string[] inputFeatureNames)
void Fit(Matrix<T> data, Vector<T> target)