private bool IsRestart(ref Solution solution, ColumnVector fitnesses, double tolX, CMAState cmaState)
{
bool isAllNaN = fitnesses.All(x => double.IsNaN(x) || double.IsInfinity(x));
if (isAllNaN)
{
return(true);
}
// Stop if the condition number of the covariance matrix
//exceeds 1014 (conditioncov).
bool isCConditionNumberTooHigh = false;
if ((Utilities.Diag(cmaState.C, x => x).ToArray().Max() / Utilities.Diag(cmaState.C, x => x).ToArray().Min()) > 1e14)
{
isCConditionNumberTooHigh = true;
}
if (isCConditionNumberTooHigh)
{
return(true);
}
if (solution.History == null)
{
solution.History = new Queue <double>();
}
//Stop if the range of the best objective function values
//of the last 10 + [30n/lambda] generations is zero
//(equalfunvalhist), or the range of these function
//values and all function values of the recent generation
//is below Tolfun= 10^-12.
solution.History.Enqueue(fitnesses[0]);
if (solution.History.Count > (10.0 + ((30.0 * cmaState.p.N) / cmaState.p.Lambda)))
{
solution.History.Dequeue();
}
bool isObjectiveFunctionRangeTooLow = (solution.History.Count == (int)(10.0 + ((30.0 * cmaState.p.N) / cmaState.p.Lambda))) &&
((solution.History.Max() == solution.History.Min()) ||
(((solution.History.Max() - solution.History.Min()) < ObjectiveFunctionTolerance) && ((fitnesses.Max() - fitnesses.Min()) < ObjectiveFunctionTolerance)));
if (isObjectiveFunctionRangeTooLow)
{
return(true);
}
// Stop if the standard deviation of the normal distribution
//is smaller than TolX in all coordinates and [sigma]pc
//(the evolution path from Eq. 2 in [3]) is smaller than
//TolX in all components. We set TolX= 10^-12*[sigma]^(0).
bool isStandardDeviationTooSmall = (Utilities.IsTrueForAll(cmaState.C, x => Math.Abs(x) < tolX) && cmaState.ps.All(x => Math.Abs(x) < tolX));
if (isStandardDeviationTooSmall)
{
return(true);
}
// Stop if adding a 0.1-standard deviation vector in
//a principal axis direction of C^(g) does not change
//<x[vector]>w^(g) (noeffectaxis)3
int ith = (cmaState.gen % cmaState.p.N);
ColumnVector tmpXmean = (cmaState.mean + 0.1 * cmaState.sigma * cmaState.B * new ColumnVector(cmaState.d));
bool isNoEffectAxis = false;
if (ith < tmpXmean.Dimension)
{
isNoEffectAxis = (tmpXmean[ith] == cmaState.mean[ith]);
}
if (isNoEffectAxis)
{
return(true);
}
// Stop if adding 0.2-standard deviation in each coordinate
//does change <x[vector]>w^(g) (noeffectcoord).
SquareMatrix testC = 0.2 * cmaState.sigma * cmaState.C;
SquareMatrix vectors;
SquareMatrix values;
Utilities.EigAlgLib(testC, out vectors, out values, 1, isGetLower: false);
ColumnVector colValues = Utilities.Diag(values, x => Math.Sqrt(Math.Abs(x)));
tmpXmean = (cmaState.mean + cmaState.sigma * vectors * colValues);
bool isNoEffectCoord = true;
for (int i = 0; i < cmaState.mean.Dimension; i++)
{
if (cmaState.mean[i] != tmpXmean[i])
{
isNoEffectCoord = false;
break;
}
}
if (isNoEffectCoord)
{
return(true);
}
return(false);
}