CMIM<T>
Models & Types ยท AiDotNet.Preprocessing.FeatureSelection.Filter.Multivariate
Conditional Mutual Information Maximization (CMIM) for feature selection.
Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> โ CMIM<T>
Implements: IFixedCountFeatureSelector<T>, ISupervisedFeatureSelector<T>
For Beginners
CMIM is cautious about redundancy. When considering a new feature, it looks at how much information it provides about the target given each already-selected feature (conditioning). It takes the worst case (minimum) and picks the feature whose worst case is best. This ensures every selected feature adds value no matter which other feature you consider.
How It Works
CMIM uses a max-min criterion to select features. For each candidate, it computes the minimum conditional mutual information with respect to all already-selected features, then selects the candidate with the maximum of these minima.
Constructors
| Constructor | Summary |
|---|---|
CMIM(int nFeaturesToSelect, int nBins, int[] columnIndices) |
Properties
| Property | Summary |
|---|---|
Vector<T> CmimScores { get; } |
|
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) |