/// <summary>
/// Selected Individuals are returned
/// </summary>
/// <param name="targetIndividuals"></param>
/// <returns></returns>
public IndividualsGroup ExecuteSelection(IndividualsGroup _targetIndividuals)
{
IndividualsGroup selectedIndividuals = new IndividualsGroup();
IndividualsGroup targetIndividuals = new IndividualsGroup(_targetIndividuals);
var elite = EliteSelection(targetIndividuals);
SelectedIndividualsIndex[0] = targetIndividuals.IndexOf(elite);
targetIndividuals.IndivisualsGroup.Remove(elite);
var roulette = RouletteSelection(targetIndividuals);
if (SelectedIndividualsIndex[0] <= targetIndividuals.IndexOf(roulette))
{
SelectedIndividualsIndex[1] = targetIndividuals.IndexOf(roulette) + 1;
}
else
{
SelectedIndividualsIndex[1] = targetIndividuals.IndexOf(roulette);
}
selectedIndividuals.AddIndivisual(elite);
selectedIndividuals.AddIndivisual(roulette);
return(selectedIndividuals);
}
private IndividualsGroup UNDX_MGG(IndividualsGroup parents, EvaluationMethod evaluationMethod)
{
UNDX_Parameters UNDXConfiguration = new UNDX_Parameters()
{
Alpha = 0.5,
Beta = 0.35,
NumberOfCrossovers = 10,
NumberOfParameters = parents.IndivisualsGroup[0].OptParameters.Length
};
UNDX undxExecutor = new UNDX(UNDXConfiguration);
MGG mggExecutor = new MGG();
// Execute Crossove with UNDX
var parameters = undxExecutor.ExecuteCrossover(parents);
var parentsIndex = undxExecutor.ParentsIndex;
// Calculate Characteristics
var fitnesses = evaluationMethod(parameters);
// Create Individuals
var offsprings = CreateIndividuals(parameters, fitnesses);
// Create New Individuals
var newIndividuals = mggExecutor.ExecuteSelection(offsprings);
// Return new population
return(UpdatePopulation(parents, (int[])parentsIndex, newIndividuals));
}
public IndividualsGroup ExecuteGenerationChange(IndividualsGroup parents)
{
if (generationChangeModel == GenerationChangeModel.UNDX_MGG)
{
return(UNDX_MGG(parents, evaluationMethod));
}
else
{
return(null);
}
}
/// <summary>
/// Create Individuals from parameters and fitnesses
/// </summary>
/// <param name="parameters"></param>
/// <param name="fitnesses"></param>
/// <returns></returns>
private IndividualsGroup CreateIndividuals(double[][] parameters, double[] fitnesses)
{
// Evaluate generated offspring
IndividualsGroup offspring = new IndividualsGroup();
var nCrossover = parameters.Length;
for (int i = 0; i < nCrossover; i++)
{
offspring.AddIndivisual(new Individual(parameters[i], fitnesses[i]));
}
return(offspring);
}
/// <summary>
/// Execute the crossover with UNDX
/// </summary>
/// <param name="parents">Individuals of previous generation</param>
/// <returns>If all crossovers are succesful, offsprings are returned, otherwise null is returned.</returns>
public double[][] ExecuteCrossover(IndividualsGroup parents)
{
Airfoil.CombinedAirfoilsGroupManager offsptingAirfoils = new Airfoil.CombinedAirfoilsGroupManager(nCrossover);
Airfoil.AirfoilManager[] selectedParents = new Airfoil.AirfoilManager[2];
General.RandomNumber.RandomNumberGenerator randomNumberGenerator = new General.RandomNumber.RandomNumberGenerator();
General.Statistics.SamplingWithoutReplacement sampling = new General.Statistics.SamplingWithoutReplacement();
// Random Airfoil Selection
uint[] pIndex = sampling.GetIndex(3, (uint)parents.NumberOfIndividuals);
ParentsIndex[0] = (int)pIndex[0];
ParentsIndex[1] = (int)pIndex[1];
// Create Optimization Parameter Vector
General.Vector p1 = new General.Vector(nParams);
General.Vector p2 = new General.Vector(nParams);
General.Vector p3 = new General.Vector(nParams);
for (int i = 0; i < nParams; i++)
{
p1[i] = parents.IndivisualsGroup[(int)pIndex[0]].OptParameters[i];
p2[i] = parents.IndivisualsGroup[(int)pIndex[1]].OptParameters[i];
p3[i] = parents.IndivisualsGroup[(int)pIndex[2]].OptParameters[i];
}
// If same parameter vector are selected, return null to retry selection.
if (p1.Equals(p2) && p1.Equals(p3))
{
return(null);
}
int n = parents.NumberOfIndividuals;
double d1 = (p2 - p1).Norm();
double d2 = (General.Vector.InnerProduct(p3 - p1, p2 - p1) / Math.Pow((p2 - p1).Norm(), 2) * (p2 - p1) - p3).Norm();
double sigma1 = alpha * d1;
double sigma2 = beta * d2 / Math.Sqrt(n);
General.Vector m = (1.0 / 2.0) * (p1 + p2);
// Create new parameter vector for offsprings
General.Vector[] offspringParameter = new General.Vector[nCrossover];
double[][] offspringParameterArray = new double[nCrossover][];
// Create Offspring's parameter vector
for (int i = 0; i < nCrossover; i++)
{
var StdOffspringPts = randomNumberGenerator.NormDistRandNumSeq(nParams);
offspringParameter[i] = new General.Vector(StdOffspringPts);
var e1 = 1 / (p2 - p1).Norm() * (p2 - p1);
var pAxisComp = General.Vector.InnerProduct(offspringParameter[i], e1);
offspringParameter[i] -= pAxisComp * e1;
offspringParameter[i] = sigma2 * offspringParameter[i];
offspringParameter[i] += sigma1 * pAxisComp * e1;
offspringParameter[i] = offspringParameter[i] + m;
offspringParameterArray[i] = offspringParameter[i].ToDoubleArray();
}
return(offspringParameterArray);
}
public IndividualsGroup(IndividualsGroup group)
{
this.nIndividuals = group.nIndividuals;
this.individuals = new List <Individual>(group.individuals);
}
