public override bool Equals(object obj)
{
Model test = obj as Model;
if (test == null)
{
return(false);
}
bool same = ClassLabels.IsEqual(test.ClassLabels);
same = same && NumberOfClasses == test.NumberOfClasses;
same = same && NumberOfSVPerClass.IsEqual(test.NumberOfSVPerClass);
if (PairwiseProbabilityA != null)
{
same = same && PairwiseProbabilityA.IsEqual(test.PairwiseProbabilityA);
}
if (PairwiseProbabilityB != null)
{
same = same && PairwiseProbabilityB.IsEqual(test.PairwiseProbabilityB);
}
same = same && Parameter.Equals(test.Parameter);
same = same && Rho.IsEqual(test.Rho);
same = same && SupportVectorCoefficients.IsEqual(test.SupportVectorCoefficients);
same = same && SupportVectorCount == test.SupportVectorCount;
same = same && SupportVectors.IsEqual(test.SupportVectors);
return(same);
}
public override void Algorithm()
{
SupportVectors.Clear();
SupportVectors.Capacity = 0;
// Save support vectors indexes
for (int i = 0; i < Solver.lagrangians.Length; ++i)
{
if (Solver.lagrangians[i] > 0)
{
SupportVectors.Add(i);
}
}
foreach (var i in SupportVectors)
{
foreach (var j in SupportVectors)
{
Lambda += Labels[i] * Labels[j] * Solver.lagrangians[i] * Solver.lagrangians[j] * KernelMatrix[i, j];
}
}
int chosen = SupportVectors.FirstOrDefault();
B = Labels[chosen] * Lambda;
foreach (var i in SupportVectors)
{
B -= Solver.lagrangians[i] * Labels[i] * KernelMatrix[chosen, i];
}
Margin = Math.Sqrt(Lambda);
}
/// <summary>
/// Creates a new object that is a copy of the current instance.
/// </summary>
/// <returns>
/// A new object that is a copy of this instance.
/// </returns>
public override object Clone()
{
var clone = new SupportVectorMachine <TKernel>(NumberOfInputs, Kernel);
clone.SupportVectors = SupportVectors.MemberwiseClone();
clone.Weights = (double[])Weights.Clone();
clone.IsProbabilistic = IsProbabilistic;
clone.Threshold = Threshold;
return(clone);
}
/// <summary>
/// Creates a new object that is a copy of the current instance.
/// </summary>
/// <returns>
/// A new object that is a copy of this instance.
/// </returns>
public virtual object Clone()
{
var clone = new SupportVectorMachine <TKernel, TInput>(NumberOfInputs, Kernel);
clone.SupportVectors = (TInput[])SupportVectors.Clone();
clone.Weights = (double[])Weights.Clone();
clone.IsProbabilistic = IsProbabilistic;
clone.Threshold = Threshold;
return(clone);
}
public override int GetHashCode()
{
return(ClassLabels.ComputeHashcode() +
NumberOfClasses.GetHashCode() +
NumberOfSVPerClass.ComputeHashcode() +
PairwiseProbabilityA.ComputeHashcode() +
PairwiseProbabilityB.ComputeHashcode() +
Parameter.GetHashCode() +
Rho.ComputeHashcode() +
SupportVectorCoefficients.ComputeHashcode() +
SupportVectorCount.GetHashCode() +
SupportVectors.ComputeHashcode());
}
public override void Algorithm()
{
SupportVectors.Clear();
SupportVectors.Capacity = 0;
// Save support vectors indexes
for (int i = 0; i < Solver.lagrangians.Length; ++i)
{
if (Solver.lagrangians[i] > 0)
{
SupportVectors.Add(i);
}
}
int indexi = -1, indexj = -1;
foreach (var i in SupportVectors)
{
if ((-1.0) / (Nu * Values.Length) < Solver.lagrangians[i] * Labels[i] && Solver.lagrangians[i] * Labels[i] < 0)
{
indexi = i;
}
if (0 < Solver.lagrangians[i] * Labels[i] && Solver.lagrangians[i] * Labels[i] < 1.0 / (Nu * Values.Length))
{
indexj = i;
}
foreach (var j in SupportVectors)
{
Lambda += Labels[i] * Labels[j] * Solver.lagrangians[i] * Solver.lagrangians[j] * KernelMatrix[i, j];
}
}
Lambda = Math.Sqrt(Lambda) / 2.0;
double left = 0, right = 0;
foreach (var k in SupportVectors)
{
left += Solver.lagrangians[k] * Labels[k] * KernelMatrix[k, indexi];
right += Solver.lagrangians[k] * Labels[k] * KernelMatrix[k, indexj];
}
B = -Lambda * (left + right);
Margin = 2 * Lambda * right + B;
}
