public IKernel Clone()
{
GaussianKernel clone = new GaussianKernel();
clone.Sigma = mSigma;
return(clone);
}
public void Select_C_and_Sigma(List <T> X, List <T> Xval, int max_iterations, out double best_C, out double best_sigma)
{
double[] C_candidates = new double[] { 0.01, 0.03, 0.1, 0.3, 1, 3, 10, 30 };
double[] sigma_candidates = new double[] { 0.01, 0.03, 0.1, 0.3, 1, 3, 10, 30 };
int C_length = C_candidates.Length;
int sigma_length = sigma_candidates.Length;
double min_prediction_error = double.MaxValue;
best_C = -1;
best_sigma = -1;
for (int i = 0; i < C_length; ++i)
{
for (int j = 0; j < sigma_length; ++j)
{
KernelSVM <T> svm = new KernelSVM <T>();
GaussianKernel kernel = svm.Kernel as GaussianKernel;
kernel.Sigma = sigma_candidates[j];
svm.C = C_candidates[i];
svm.MaxSolverIteration = max_iterations;
svm.Train(X);
double prediction_error = svm.GetPredictionError(Xval);
if (min_prediction_error > prediction_error)
{
min_prediction_error = prediction_error;
best_C = C_candidates[i];
best_sigma = sigma_candidates[j];
}
}
}
}
