/// <summary>
/// The input vector cannot be sparse. Method returns 0 if class does not exist
/// </summary>
/// <param name="row"></param>
/// <returns></returns>
public double MahalanobisDistance(PatternTools.sparseMatrixRow row)
{
//Term 1 (xi - mi)Transposed
//Term 2 Inverse Cov Matrix
//Term 3 (xi - mi)
//M Squared = Math.Sqrt (t1 * t2 * t3)
//Eliminate unstable dims
if (classLableClassPackageDic.ContainsKey(row.Lable))
{
Matrix t1 = new CSML.Matrix(row.Values.ToArray()) - classLableClassPackageDic[row.Lable].MatrixMeanVector;
t1 = t1.Transpose();
Matrix t2 = classLableClassPackageDic[row.Lable].InverseCovarianceMatrix;
Matrix t3 = new CSML.Matrix(row.Values.ToArray()) - classLableClassPackageDic[row.Lable].MatrixMeanVector;
Matrix product = t1 * t2 * t3;
double det = Math.Abs(product.Determinant().Re);
return(Math.Pow(det, 0.5));
}
else
{
return(0);
}
}
private double[] Unsparse(PatternTools.sparseMatrixRow r)
{
//Lets unsparse the point
double[] point = new double[r.Dims.Count];
for (int i = 0; i < point.Length; i++)
{
point[i] = r.Values[i];
}
return(point);
}
//----------------------------
public int SimpleClassification(PatternTools.sparseMatrixRow r)
{
return(SimpleClassification(Unsparse(r)));
}
