MIFS<T>

Models & Types ยท AiDotNet.Preprocessing.FeatureSelection.Filter.Multivariate

Mutual Information Feature Selection (MIFS) with redundancy penalty.

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

Implements: IFixedCountFeatureSelector<T>, ISupervisedFeatureSelector<T>

For Beginners

MIFS greedily picks features that tell you a lot about the target. But it also subtracts a penalty if the new feature overlaps with what you already know from selected features. The beta parameter controls how harsh this penalty is: 0 = no penalty (just pick high MI), 1 = full penalty.

How It Works

MIFS iteratively selects features that have high mutual information with the target while penalizing redundancy with already-selected features. The beta parameter controls the strength of the redundancy penalty.

Constructors

Constructor Summary
MIFS(int nFeaturesToSelect, int nBins, double beta, int[] columnIndices)

Properties

Property Summary
Vector<T> RelevanceScores { 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)