public void Reset()
{
PSigma.Clear();
PC.Clear();
HSigma = 0;
Sigma = initSigma;
}
public void UpdateEvolutionPaths(Matrix <double> eigenvectors, Matrix <double> invertedSqrtEigenvalues,
AdaptationParameters adaptationParameters, SelectionParameters selectionParameters,
long FFE, Vector <double> means, Vector <double> previousMeans)
{
int problemSize = PSigma.Count;
Matrix <double> invertedSqrtCovarianceMatrix = eigenvectors * invertedSqrtEigenvalues * eigenvectors.Transpose();
Vector <double> meanDifferenceDivBySigma = (means - previousMeans) / Sigma;
PSigma *= 1 - adaptationParameters.CSigma;
PSigma += invertedSqrtCovarianceMatrix * PathUpdateCoefficient(adaptationParameters.CSigma, selectionParameters.MuEff, meanDifferenceDivBySigma);
HSigma = Convert.ToInt32(PSigma.DotProduct(PSigma) / (1.0 - Math.Pow(1.0 - adaptationParameters.CSigma,
2.0 * FFE / selectionParameters.Lambda)) / problemSize
< 2.0 + 4.0 / (problemSize + 1.0));
PC *= 1.0 - adaptationParameters.CC;
PC += HSigma * PathUpdateCoefficient(adaptationParameters.CC, selectionParameters.MuEff, meanDifferenceDivBySigma);
}
public void AdaptStepSize(AdaptationParameters adaptationParameters)
{
Sigma *= Math.Exp(adaptationParameters.CSigma / adaptationParameters.Damping * (PSigma.L2Norm() / adaptationParameters.ChiN - 1.0));
}
