public DimReduction(double[][] trainingMatrix)
{
//Create the Principal Component Analysis
_pca = new ModifedPca(trainingMatrix);
_pca.Compute();
PrintEngine.printList(_pca.Eigenvalues.ToList(), Form1.MainFolderName + "eigvalues.txt");
}
/// <summary>
/// Constructs a new Principal Component Analysis from a Covariance matrix.
/// </summary>
///
/// <remarks>
/// This method may be more suitable to high dimensional problems in which
/// the original data matrix may not fit in memory but the covariance matrix
/// will.</remarks>
///
/// <param name="mean">The mean vector for the source data.</param>
/// <param name="covariance">The covariance matrix of the data.</param>
///
public static ModifedPca FromCovarianceMatrix(double[] mean, double[,] covariance)
{
if (mean == null)
{
throw new ArgumentNullException("mean");
}
if (covariance == null)
{
throw new ArgumentNullException("covariance");
}
if (covariance.GetLength(0) != covariance.GetLength(1))
{
throw new NonSymmetricMatrixException("Covariance matrix must be symmetric");
}
ModifedPca pca = new ModifedPca();
pca.columnMeans = mean;
pca.covarianceMatrix = covariance;
pca.analysisMethod = AnalysisMethod.Center;
pca.onlyCovarianceMatrixAvailable = true;
pca.sourceDimensions = covariance.GetLength(0);
return(pca);
}
/// <summary>
/// Constructs a new Principal Component Analysis from a Correlation matrix.
/// </summary>
///
/// <remarks>
/// This method may be more suitable to high dimensional problems in which
/// the original data matrix may not fit in memory but the covariance matrix
/// will.</remarks>
///
/// <param name="mean">The mean vector for the source data.</param>
/// <param name="stdDev">The standard deviation vectors for the source data.</param>
/// <param name="correlation">The correlation matrix of the data.</param>
///
public static ModifedPca FromCorrelationMatrix(double[] mean, double[] stdDev, double[,] correlation)
{
if (mean == null)
{
throw new ArgumentNullException("mean");
}
if (stdDev == null)
{
throw new ArgumentNullException("stdDev");
}
if (correlation == null)
{
throw new ArgumentNullException("correlation");
}
if (correlation.GetLength(0) != correlation.GetLength(1))
{
throw new NonSymmetricMatrixException("Correlation matrix must be symmetric");
}
ModifedPca pca = new ModifedPca();
pca.columnMeans = mean;
pca.columnStdDev = stdDev;
pca.covarianceMatrix = correlation;
pca.analysisMethod = AnalysisMethod.Standardize;
pca.onlyCovarianceMatrixAvailable = true;
pca.sourceDimensions = correlation.GetLength(0);
return(pca);
}
/// <summary>
/// Creates a principal component representation.
/// </summary>
///
/// <param name="analysis">The analysis to which this component belongs.</param>
/// <param name="index">The component index.</param>
///
internal PrincipalComponent(ModifedPca analysis, int index)
{
this.index = index;
this._modifedPca = analysis;
}
public static void originalFeatureRole(ModifedPca nodePca)
{
//TO DO: baced on http://venom.cs.utsa.edu/dmz/techrep/2007/CS-TR-2007-011.pdf
}
