public CrossPRESSEvaluator(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
: base(spectralRegions,numFactors,groupingStrategy,preprocessOptions,analysis)
{
}
public CrossPRESSEvaluator(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
: base(spectralRegions, numFactors, groupingStrategy, preprocessOptions, analysis)
{
}
public CrossPredictedXResidualsEvaluator(
int numberOfPoints,
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
: base(spectralRegions, numFactors, groupingStrategy, preprocessOptions, analysis)
{
_numberOfPoints = numberOfPoints;
}
public CrossPredictedYEvaluator(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis,
IMatrix YCrossValidationPrediction
)
: base(spectralRegions, numFactors, groupingStrategy, preprocessOptions, analysis)
{
_YCrossValidationPrediction = YCrossValidationPrediction;
}
public CrossValidationWorker(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
{
_spectralRegions = spectralRegions;
_numFactors = numFactors;
_groupingStrategy = groupingStrategy;
_preprocessOptions = preprocessOptions;
_analysis = analysis;
}
public CrossValidationWorker(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
{
_spectralRegions = spectralRegions;
_numFactors = numFactors;
_groupingStrategy = groupingStrategy;
_preprocessOptions = preprocessOptions;
_analysis = analysis;
}
public CrossPredictedYEvaluator(
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis,
IMatrix YCrossValidationPrediction
)
: base(spectralRegions,numFactors,groupingStrategy,preprocessOptions,analysis)
{
_YCrossValidationPrediction = YCrossValidationPrediction;
}
public CrossPredictedXResidualsEvaluator(
int numberOfPoints,
int[] spectralRegions,
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression analysis
)
: base(spectralRegions,numFactors,groupingStrategy,preprocessOptions,analysis)
{
_numberOfPoints = numberOfPoints;
}
/// <summary>
/// Calculates the spectral residuals obtained from cross validation.
/// </summary>
/// <param name="xOfX">The spectral wavelength values corresponding to the spectral bins.</param>
/// <param name="X">Matrix of spectra (a spectrum = a row in the matrix).</param>
/// <param name="Y">Matrix of y values (e.g. concentrations).</param>
/// <param name="numFactors">Number of factors used for calculation.</param>
/// <param name="groupingStrategy">The strategy how to group the spectra for cross prediction.</param>
/// <param name="preprocessOptions">Information how to preprocess the data.</param>
/// <param name="regress">The type of regression (e.g. PCR, PLS1, PLS2) provided as an empty regression object.</param>
/// <param name="crossXResiduals">Returns the matrix of spectral residuals</param>
/// <returns>Mean number of spectra used for prediction.</returns>
public static double GetCrossXResiduals(
IROVector xOfX,
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression regress,
out IROMatrix crossXResiduals
)
{
return GetCrossXResiduals(
SpectralPreprocessingOptions.IdentifyRegions(xOfX),
X, // matrix of spectra (a spectra is a row of this matrix)
Y, // matrix of concentrations (a mixture is a row of this matrix)
numFactors,
groupingStrategy,
preprocessOptions,
regress,
out crossXResiduals);
}
/// <summary>
///
/// </summary>
/// <param name="mcalib"></param>
/// <param name="groupingStrategy"></param>
/// <param name="preprocessOptions"></param>
/// <param name="xOfX"></param>
/// <param name="matrixX">Matrix of horizontal spectra, centered and preprocessed.</param>
/// <param name="matrixY">Matrix of concentrations, centered.</param>
/// <param name="numberOfFactors"></param>
/// <param name="predictedY"></param>
/// <param name="spectralResiduals"></param>
public virtual void CalculateCrossPredictedY(
IMultivariateCalibrationModel mcalib,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
IROVector xOfX,
IMatrix matrixX,
IMatrix matrixY,
int numberOfFactors,
IMatrix predictedY,
IMatrix spectralResiduals)
{
MultivariateRegression.GetCrossYPredicted(xOfX,
matrixX, matrixY, numberOfFactors, groupingStrategy, preprocessOptions,
this.CreateNewRegressionObject(),
predictedY);
}
/// <summary>
/// Calculates the spectral residuals obtained from cross validation.
/// </summary>
/// <param name="spectralRegions">Array of ascending indices representing the starting indices of spectral regions.</param>
/// <param name="X">Matrix of spectra (a spectrum = a row in the matrix).</param>
/// <param name="Y">Matrix of y values (e.g. concentrations).</param>
/// <param name="numFactors">Number of factors used for calculation.</param>
/// <param name="groupingStrategy">The strategy how to group the spectra for cross prediction.</param>
/// <param name="preprocessOptions">Information how to preprocess the data.</param>
/// <param name="regress">The type of regression (e.g. PCR, PLS1, PLS2) provided as an empty regression object.</param>
/// <param name="crossXResiduals">Returns the matrix of spectral residuals</param>
/// <returns>Mean number of spectra used for prediction.</returns>
public static double GetCrossXResiduals(
int[] spectralRegions,
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression regress,
out IROMatrix crossXResiduals
)
{
CrossPredictedXResidualsEvaluator worker = new CrossPredictedXResidualsEvaluator(X.Rows,spectralRegions,numFactors,groupingStrategy,preprocessOptions,regress);
double result = CrossValidationIteration(X,Y,groupingStrategy,new CrossValidationIterationFunction(worker.EhCrossValidationWorker));
crossXResiduals = worker.XCrossResiduals;
return result;
}
/// <summary>
/// Calculates the cross predicted y values.
/// </summary>
/// <param name="xOfX">The spectral wavelength values corresponding to the spectral bins.</param>
/// <param name="X">Matrix of spectra (a spectrum = a row in the matrix).</param>
/// <param name="Y">Matrix of y values (e.g. concentrations).</param>
/// <param name="numFactors">Number of factors used for calculation.</param>
/// <param name="groupingStrategy">The strategy how to group the spectra for cross prediction.</param>
/// <param name="preprocessOptions">Information how to preprocess the data.</param>
/// <param name="regress">The type of regression (e.g. PCR, PLS1, PLS2) provided as an empty regression object.</param>
/// <param name="yCrossPredicted">Matrix of cross predicted y values. Must be of same dimension as the Y matrix.</param>
/// <returns>Mean number of spectra used for cross prediction.</returns>
public static double GetCrossYPredicted(
IROVector xOfX,
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression regress,
IMatrix yCrossPredicted // vertical value of PRESS values for the cross validation
)
{
return GetCrossYPredicted(
SpectralPreprocessingOptions.IdentifyRegions(xOfX),
X, // matrix of spectra (a spectra is a row of this matrix)
Y, // matrix of concentrations (a mixture is a row of this matrix)
numFactors,
groupingStrategy,
preprocessOptions,
regress,
yCrossPredicted );
}
/// <summary>
/// Calculates the cross predicted y values.
/// </summary>
/// <param name="spectralRegions">Array of ascending indices representing the starting indices of spectral regions.</param>
/// <param name="X">Matrix of spectra (a spectrum = a row in the matrix).</param>
/// <param name="Y">Matrix of y values (e.g. concentrations).</param>
/// <param name="numFactors">Number of factors used for calculation.</param>
/// <param name="groupingStrategy">The strategy how to group the spectra for cross prediction.</param>
/// <param name="preprocessOptions">Information how to preprocess the data.</param>
/// <param name="regress">The type of regression (e.g. PCR, PLS1, PLS2) provided as an empty regression object.</param>
/// <param name="yCrossPredicted">Matrix of cross predicted y values. Must be of same dimension as the Y matrix.</param>
/// <returns>Mean number of spectra used for cross prediction.</returns>
public static double GetCrossYPredicted(
int[] spectralRegions,
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression regress,
IMatrix yCrossPredicted // vertical value of PRESS values for the cross validation
)
{
CrossPredictedYEvaluator worker = new CrossPredictedYEvaluator(spectralRegions,numFactors,groupingStrategy,preprocessOptions,regress,yCrossPredicted);
double result = CrossValidationIteration(X,Y,groupingStrategy,new CrossValidationIterationFunction(worker.EhYCrossPredicted));
return result;
}
/// <summary>
/// Get the cross predicted error sum of squares for the number of factors=0...numFactors.
/// </summary>
/// <param name="spectralRegions">Array of ascending indices representing the starting indices of spectral regions.</param>
/// <param name="X">Matrix of spectra (a spectrum = a row in the matrix).</param>
/// <param name="Y">Matrix of y values (e.g. concentrations).</param>
/// <param name="numFactors">Maximum number of factors to calculate the cross PRESS for.</param>
/// <param name="groupingStrategy">The strategy how to group the spectra for cross prediction.</param>
/// <param name="preprocessOptions">Information how to preprocess the data.</param>
/// <param name="regress">The type of regression (e.g. PCR, PLS1, PLS2) provided as an empty regression object.</param>
/// <param name="crossPRESS">The vector of CROSS press values. Note that this vector has the length numFactor+1.</param>
/// <returns>The mean number of spectra used for prediction.</returns>
public static double GetCrossPRESS(
int[] spectralRegions,
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
int numFactors,
ICrossValidationGroupingStrategy groupingStrategy,
SpectralPreprocessingOptions preprocessOptions,
MultivariateRegression regress,
out IROVector crossPRESS // vertical value of PRESS values for the cross validation
)
{
CrossPRESSEvaluator worker = new CrossPRESSEvaluator(spectralRegions,numFactors,groupingStrategy,preprocessOptions,regress);
double result = CrossValidationIteration(X,Y,groupingStrategy,new CrossValidationIterationFunction(worker.EhCrossPRESS));
crossPRESS = VectorMath.ToROVector(worker.CrossPRESS,worker.NumberOfFactors+1);
return result;
}
/// <summary>
/// This function separates the spectra into a bunch of spectra used for calibration and the rest of spectra
/// used for prediction. This separation is repeated until all spectra are used exactly one time for prediction.
/// </summary>
/// <param name="X">Matrix of spectra (horizontal oriented).</param>
/// <param name="Y">Matrix of y values.</param>
/// <param name="groupingStrategy">The strategy how to separate the spectra into the calibration and prediction spectra.</param>
/// <param name="crossFunction">The function that is called for each separation.</param>
/// <returns>The mean number of spectra that was used for prediction.</returns>
public static double CrossValidationIteration(
IROMatrix X, // matrix of spectra (a spectra is a row of this matrix)
IROMatrix Y, // matrix of concentrations (a mixture is a row of this matrix)
ICrossValidationGroupingStrategy groupingStrategy,
CrossValidationIterationFunction crossFunction
)
{
// int[][] groups = bExcludeGroups ? new ExcludeGroupsGroupingStrategy().Group(Y) : new ExcludeSingleMeasurementsGroupingStrategy().Group(Y);
int[][] groups = groupingStrategy.Group(Y);
IMatrix XX=null;
IMatrix YY=null;
IMatrix XU=null;
IMatrix YU=null;
for(int nGroup=0 ,prevNumExcludedSpectra = int.MinValue ;nGroup < groups.Length;nGroup++)
{
int[] spectralGroup = groups[nGroup];
int numberOfExcludedSpectraOfGroup = spectralGroup.Length;
if(prevNumExcludedSpectra != numberOfExcludedSpectraOfGroup)
{
XX = new MatrixMath.BEMatrix(X.Rows-numberOfExcludedSpectraOfGroup,X.Columns);
YY = new MatrixMath.BEMatrix(Y.Rows-numberOfExcludedSpectraOfGroup,Y.Columns);
XU = new MatrixMath.BEMatrix(numberOfExcludedSpectraOfGroup,X.Columns);
YU = new MatrixMath.BEMatrix(numberOfExcludedSpectraOfGroup,Y.Columns);
prevNumExcludedSpectra = numberOfExcludedSpectraOfGroup;
}
// build a new x and y matrix with the group information
// fill XX and YY with values
for(int i=0,j=0;i<X.Rows;i++)
{
if(Array.IndexOf(spectralGroup,i)>=0) // if spectral group contains i
continue; // Exclude this row from the spectra
MatrixMath.SetRow(X,i,XX,j);
MatrixMath.SetRow(Y,i,YY,j);
j++;
}
// fill XU (unknown spectra) with values
for(int i=0;i<spectralGroup.Length;i++)
{
int j = spectralGroup[i];
MatrixMath.SetRow(X,j,XU,i); // x-unkown (unknown spectra)
MatrixMath.SetRow(Y,j,YU,i); // y-unkown (unknown concentration)
}
// now do the analysis
crossFunction(spectralGroup,XX,YY,XU,YU);
} // for all groups
// calculate the mean number of excluded spectras
return ((double)X.Rows)/groups.Length;
}
