private void init(double[,] inputs, int[] outputs)
{
// Gets the number of classes
int startingClass = outputs.Min();
this.classes = outputs.Max() - startingClass + 1;
// Store the original data
this.source = inputs;
this.outputs = outputs;
this.dimension = inputs.GetLength(1);
// Creates simple structures to hold information later
this.classCount = new int[classes];
this.classMeans = new double[classes][];
this.classStdDevs = new double[classes][];
this.classScatter = new double[classes][, ];
this.projectedMeans = new double[classes][];
// Creates the object-oriented structure to hold information about the classes
DiscriminantAnalysisClass[] collection = new DiscriminantAnalysisClass[classes];
for (int i = 0; i < classes; i++)
{
collection[i] = new DiscriminantAnalysisClass(this, i, startingClass + i);
}
this.classCollection = new DiscriminantAnalysisClassCollection(collection);
}
//---------------------------------------------
#region Constructor
/// <summary>
/// Constructs a new Linear Discriminant Analysis object.
/// </summary>
/// <param name="inputs">The source data to perform analysis. The matrix should contain
/// variables as columns and observations of each variable as rows.</param>
/// <param name="outputs">The labels for each observation row in the input matrix.</param>
public LinearDiscriminantAnalysis(double[,] inputs, int[] outputs)
{
// Initial argument checking
if (inputs == null)
{
throw new ArgumentNullException("inputs");
}
if (outputs == null)
{
throw new ArgumentNullException("outputs");
}
if (inputs.GetLength(0) != outputs.Length)
{
throw new ArgumentException("The number of rows in the input array must match the number of given outputs.");
}
// Gets the number of classes
int startingClass = outputs.Min();
this.classes = outputs.Max() - startingClass + 1;
// Store the original data
this.source = inputs;
this.outputs = outputs;
this.dimension = inputs.GetLength(1);
// Creates simple structures to hold information later
this.classCount = new int[classes];
this.classMeans = new double[classes][];
this.classStdDevs = new double[classes][];
this.classScatter = new double[classes][, ];
this.projectedMeans = new double[classes][];
// Creates the object-oriented structure to hold information about the classes
DiscriminantAnalysisClass[] collection = new DiscriminantAnalysisClass[classes];
for (int i = 0; i < classes; i++)
{
collection[i] = new DiscriminantAnalysisClass(this, i, startingClass + i);
}
this.classCollection = new DiscriminantAnalysisClassCollection(collection);
}
/// <summary>
/// Creates additional information about principal components.
/// </summary>
///
protected void CreateDiscriminants()
{
if (classCollection == null)
{
// Creates simple structures to hold information later
this.classCount = new int[NumberOfOutputs];
this.classMeans = new double[NumberOfOutputs][];
this.classStdDevs = new double[NumberOfOutputs][];
this.classScatter = new double[NumberOfOutputs][][];
this.projectedMeans = new double[NumberOfOutputs][];
// Creates the object-oriented structure to hold information about the classes
var collection = new DiscriminantAnalysisClass[NumberOfOutputs];
for (int i = 0; i < collection.Length; i++)
{
collection[i] = new DiscriminantAnalysisClass(this, i, i);
}
this.classCollection = new DiscriminantAnalysisClassCollection(collection);
}
int numDiscriminants = eigenvalues.Length;
discriminantProportions = new double[numDiscriminants];
discriminantCumulative = new double[numDiscriminants];
// Calculate total scatter matrix
St = Sw.Add(Sb);
// Calculate proportions
double sum = 0.0;
for (int i = 0; i < numDiscriminants; i++)
{
sum += System.Math.Abs(eigenvalues[i]);
}
sum = (sum == 0) ? 0 : (1.0 / sum);
for (int i = 0; i < numDiscriminants; i++)
{
discriminantProportions[i] = System.Math.Abs(eigenvalues[i]) * sum;
}
// Calculate cumulative proportions
if (numDiscriminants > 0)
{
this.discriminantCumulative[0] = this.discriminantProportions[0];
for (int i = 1; i < this.discriminantCumulative.Length; i++)
{
this.discriminantCumulative[i] = this.discriminantCumulative[i - 1] + this.discriminantProportions[i];
}
}
// Creates the object-oriented structure to hold the linear discriminants
var discriminants = new Discriminant[numDiscriminants];
for (int i = 0; i < discriminants.Length; i++)
{
discriminants[i] = new Discriminant(this, i);
}
this.discriminantCollection = new DiscriminantCollection(discriminants);
}