SpikeAndSlabSelector<T>

Models & Types · AiDotNet.Preprocessing.FeatureSelection.Bayesian

Spike-and-Slab Feature Selection using Bayesian variable selection.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → SpikeAndSlabSelector<T>

Implements: IFixedCountFeatureSelector<T>, ISupervisedFeatureSelector<T>

For Beginners

Imagine a spike-and-slab as two possibilities for each feature: either it has no effect (spike at zero) or it has some effect (spread out slab). The method estimates which possibility is more likely for each feature. If the slab is more likely, the feature is important.

How It Works

Implements the spike-and-slab prior model where each feature coefficient has a mixture prior: a “spike” (concentrated at zero) and a “slab” (diffuse). Features with high probability of being in the slab are selected.

Constructors

Constructor Summary
SpikeAndSlabSelector(int nFeaturesToSelect, double slabVariance, double spikeVariance, double priorInclusionProbability, int nIterations, int[] columnIndices)

Properties

Property Summary
Vector<T> InclusionProbabilities { get; }
bool SupportsInverseTransform { 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)