public IList <GAParametersChromosome> Execute(IList <GAParametersChromosome> parents)
{
GAParametersChromosome par1 = (GAParametersChromosome)parents[0].Copy();
GAParametersChromosome par2 = (GAParametersChromosome)parents[1].Copy();
IList <GAParametersChromosome> result = new List <GAParametersChromosome>();
if (RandomNumberGeneratorProvider.Instance.NextDouble() > _crossoverProbability)
{
par1.IsEvaluated = true;
par2.IsEvaluated = true;
result.Add(par1);
result.Add(par2);
return(result);
}
Tuple <GAParameters, GAParameters> child = GAParameters.Cross(par1.GetValue(0), par2.GetValue(0));
par1.SetValue(0, child.Item1);
par1.IsEvaluated = false;
par2.SetValue(0, child.Item2);
par2.IsEvaluated = false;
result.Add(par1);
return(result);
}
public GAParametersChromosome Execute(GAParametersChromosome chromosome)
{
GAParametersChromosome child = (GAParametersChromosome)chromosome.Copy();
child.IsEvaluated = true;
if (RandomNumberGeneratorProvider.Instance.NextDouble() <= _mutationProbability)
{
child.SetValue(0, GAParameters.Mutate(child.GetValue(0)));
child.IsEvaluated = false;
}
return(child);
}
public static GAParameters CreateRandom()
{
Random r = RandomNumberGeneratorProvider.Instance;
GAParameters g = new GAParameters();
g.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
g.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
g.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
g.MutationRateIndex = r.Next(0, MutationRate.Length);
g.ElitismRateIndex = r.Next(0, ElitismRate.Length);
g.GenerationGapIndex = r.Next(0, GenerationGap.Length);
g.GAVariantIndex = r.Next(0, GAVariant.Length);
g.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
g.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
g.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
g.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
g.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
while (!g.IsValid())
{
g.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
g.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
g.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
g.MutationRateIndex = r.Next(0, MutationRate.Length);
g.ElitismRateIndex = r.Next(0, ElitismRate.Length);
g.GenerationGapIndex = r.Next(0, GenerationGap.Length);
g.GAVariantIndex = r.Next(0, GAVariant.Length);
g.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
g.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
g.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
g.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
g.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
return(g);
}
public static Tuple <GAParameters, GAParameters> Cross(GAParameters g1, GAParameters g2)
{
var r = RandomNumberGeneratorProvider.Instance;
var prob = r.NextDouble();
if (prob < 0.05)
{
if (r.NextDouble() < 0.5)
{
g1.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
else
{
g2.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
}
else if (prob < 0.1)
{
if (r.NextDouble() < 0.5)
{
g1.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
else
{
g2.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
}
else if (prob < 0.15)
{
if (r.NextDouble() < 0.5)
{
g1.MutationRateIndex = r.Next(0, MutationRate.Length);
}
else
{
g2.MutationRateIndex = r.Next(0, MutationRate.Length);
}
}
else if (prob < 0.20)
{
if (r.NextDouble() < 0.5)
{
g1.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
else
{
g2.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
}
else if (prob < 0.25)
{
if (r.NextDouble() < 0.5)
{
g1.GAVariantIndex = r.Next(0, GAVariant.Length);
}
else
{
g2.GAVariantIndex = r.Next(0, GAVariant.Length);
}
}
else if (prob < 0.30)
{
if (r.NextDouble() < 0.5)
{
g1.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
else
{
g2.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
}
else if (prob < 0.35)
{
if (r.NextDouble() < 0.5)
{
g1.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
else
{
g2.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
}
else if (prob < 0.40)
{
if (r.NextDouble() < 0.5)
{
g1.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
else
{
g2.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
}
else if (prob < 0.45)
{
if (r.NextDouble() < 0.5)
{
g1.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
else
{
g2.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
else
{
g2.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
else
{
g2.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
else
{
g2.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
}
while (!g1.IsValid() && !g2.IsValid())
{
if (prob < 0.05)
{
if (r.NextDouble() < 0.5)
{
g1.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
else
{
g2.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
}
else if (prob < 0.1)
{
if (r.NextDouble() < 0.5)
{
g1.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
else
{
g2.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
}
else if (prob < 0.15)
{
if (r.NextDouble() < 0.5)
{
g1.MutationRateIndex = r.Next(0, MutationRate.Length);
}
else
{
g2.MutationRateIndex = r.Next(0, MutationRate.Length);
}
}
else if (prob < 0.20)
{
if (r.NextDouble() < 0.5)
{
g1.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
else
{
g2.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
}
else if (prob < 0.25)
{
if (r.NextDouble() < 0.5)
{
g1.GAVariantIndex = r.Next(0, GAVariant.Length);
}
else
{
g2.GAVariantIndex = r.Next(0, GAVariant.Length);
}
}
else if (prob < 0.30)
{
if (r.NextDouble() < 0.5)
{
g1.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
else
{
g2.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
}
else if (prob < 0.35)
{
if (r.NextDouble() < 0.5)
{
g1.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
else
{
g2.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
}
else if (prob < 0.40)
{
if (r.NextDouble() < 0.5)
{
g1.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
else
{
g2.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
}
else if (prob < 0.45)
{
if (r.NextDouble() < 0.5)
{
g1.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
else
{
g2.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
else
{
g2.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
else
{
g2.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
}
else if (prob < 0.50)
{
if (r.NextDouble() < 0.5)
{
g1.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
else
{
g2.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
}
}
return(new Tuple <GAParameters, GAParameters>(g1, g2));
}
public static GAParameters Mutate(GAParameters parameters)
{
var g = parameters.Copy();
var r = RandomNumberGeneratorProvider.Instance;
var idx = r.Next(0, 13);
if (idx == 0)
{
g.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
else if (idx == 1)
{
g.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
else if (idx == 2)
{
g.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
else if (idx == 3)
{
g.MutationRateIndex = r.Next(0, MutationRate.Length);
}
else if (idx == 4)
{
g.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
else if (idx == 5)
{
g.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
else if (idx == 6)
{
g.GAVariantIndex = r.Next(0, GAVariant.Length);
}
else if (idx == 7)
{
g.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
else if (idx == 8)
{
g.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
else if (idx == 9)
{
g.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
else if (idx == 10)
{
g.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
else if (idx == 11)
{
g.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
while (!g.IsValid())
{
if (idx == 0)
{
g.ChromosomeRepresentationIndex = r.Next(0, ChromosomeRepresentation.Length);
}
else if (idx == 1)
{
g.PopulationSizeIndex = r.Next(0, PopulationSize.Length);
}
else if (idx == 2)
{
g.CrossoverRateIndex = r.Next(0, CrossoverRate.Length);
}
else if (idx == 3)
{
g.MutationRateIndex = r.Next(0, MutationRate.Length);
}
else if (idx == 4)
{
g.ElitismRateIndex = r.Next(0, ElitismRate.Length);
}
else if (idx == 5)
{
g.GenerationGapIndex = r.Next(0, GenerationGap.Length);
}
else if (idx == 6)
{
g.GAVariantIndex = r.Next(0, GAVariant.Length);
}
else if (idx == 7)
{
g.SelectionOperatorIndex = r.Next(0, SelectionOperator.Length);
}
else if (idx == 8)
{
g.CrossoverOperatorIndex = r.Next(0, CrossoverOperator.Length);
}
else if (idx == 9)
{
g.MutationOperatorIndex = r.Next(0, MutationOperator.Length);
}
else if (idx == 10)
{
g.TournamentSizeIndex = r.Next(0, TournamentSize.Length);
}
else if (idx == 11)
{
g.SelectivePressureIndex = r.Next(0, SelectivePressure.Length);
}
}
return(g);
}
public void Evaluate(GAParametersChromosome chromosome)
{
// if (chromosome.IsEvaluated) return;
_evaluations++;
GAParameters parameters = chromosome.GetValue(0);
_function.Clear();
int iters = 10;
int childEvaluations = 10000;
int maxGenerations = 100;
double[] childGAFitnesses = new double[iters];
var chromosomeRepresentation = GAParameters.ChromosomeRepresentation[parameters.ChromosomeRepresentationIndex];
for (int i = 0; i < iters; i++)
{
if (chromosomeRepresentation == GAParameters.ChromosomeRepresentationType.Binary)
{
IProblem <BinaryChromosome> problem = new FunctionMinimizationBinaryProblem(_function, _functionDimension, 4, -50, 150);
int populationSize = GAParameters.PopulationSize[parameters.PopulationSizeIndex];
double crossoverRate = GAParameters.CrossoverRate[parameters.CrossoverRateIndex];
ICrossoverOperator <BinaryChromosome> crossoverOperator;
ISelectionOperator <BinaryChromosome> selectionOperator;
int tournamentSize = GAParameters.TournamentSize[parameters.TournamentSizeIndex];
double selectivePressure = GAParameters.SelectivePressure[parameters.SelectivePressureIndex];
IMutationOperator <BinaryChromosome> mutationOperator;
double mutationRate = GAParameters.MutationRate[parameters.MutationRateIndex];
GeneticAlgorithmVariant variant = GAParameters.GAVariant[parameters.GAVariantIndex];
double elitismRate = GAParameters.ElitismRate[parameters.ElitismRateIndex];
double generationGap = GAParameters.GenerationGap[parameters.GenerationGapIndex];
var crossOperatorType = GAParameters.CrossoverOperator[parameters.CrossoverOperatorIndex];
if (crossOperatorType == GAParameters.CrossoverOperatorType.OnePoint)
{
crossoverOperator = new OnePointCrossoverOperator(crossoverRate);
}
else
{
crossoverOperator = new TwoPointCrossoverOperator(crossoverRate);
}
var mutationOperatorType = GAParameters.MutationOperator[parameters.MutationOperatorIndex];
if (mutationOperatorType == GAParameters.MutationOperatorType.BinaryBoundary)
{
mutationOperator = new BinaryBoundaryMutationOperator(mutationRate);
}
else if (mutationOperatorType == GAParameters.MutationOperatorType.BitFlip)
{
mutationOperator = new BitFlipMutationOperator(mutationRate);
}
else
{
mutationOperator = new SimpleMutationOperator(mutationRate);
}
if (GAParameters.SelectionOperator[parameters.SelectionOperatorIndex] == GAParameters.SelectionOperatorType.Tournament)
{
selectionOperator = new TournamentSelectionOperator <BinaryChromosome>(tournamentSize, problem);
}
else
{
selectionOperator = new RouletteWheelSelectionOperator <BinaryChromosome>(selectivePressure, problem);
}
IGeneticAlgorithm <BinaryChromosome> ga;
if (variant == GeneticAlgorithmVariant.Generational)
{
ga = new GenerationalGeneticAlgorithm <BinaryChromosome>(
problem,
populationSize,
selectionOperator,
crossoverOperator,
mutationOperator,
new CompositeTerminationCondition(
new MaxEvaluationsTerminationCondition(childEvaluations),
new MaxGenerationsTerminationCondition(maxGenerations)
),
elitismRate
);
}
else
{
ga = new SteadyStateGeneticAlgorithm <BinaryChromosome>(
problem,
populationSize,
selectionOperator,
crossoverOperator,
mutationOperator,
new CompositeTerminationCondition(
new MaxEvaluationsTerminationCondition(childEvaluations),
new MaxGenerationsTerminationCondition(maxGenerations)
),
generationGap,
new WorstFitnessReplacementOperator <BinaryChromosome>(problem)
);
}
// new GeneticAlgorithmRunner<BinaryChromosome>(ga).Run();
ga.Run();
childGAFitnesses[i] = ga.BestIndividual.Fitness;
}
else
{
IProblem <FloatingPointChromosome> problem = new FunctionMinimizationFloatingPointProblem(_function, _functionDimension, -50, 150);
int populationSize = GAParameters.PopulationSize[parameters.PopulationSizeIndex];
double crossoverRate = GAParameters.CrossoverRate[parameters.CrossoverRateIndex];
ICrossoverOperator <FloatingPointChromosome> crossoverOperator;
ISelectionOperator <FloatingPointChromosome> selectionOperator;
int tournamentSize = GAParameters.TournamentSize[parameters.TournamentSizeIndex];
double selectivePressure = GAParameters.SelectivePressure[parameters.SelectivePressureIndex];
IMutationOperator <FloatingPointChromosome> mutationOperator;
double mutationRate = GAParameters.MutationRate[parameters.MutationRateIndex];
GeneticAlgorithmVariant variant = GAParameters.GAVariant[parameters.GAVariantIndex];
double elitismRate = GAParameters.ElitismRate[parameters.ElitismRateIndex];
double generationGap = GAParameters.GenerationGap[parameters.GenerationGapIndex];
var crossOperatorType = GAParameters.CrossoverOperator[parameters.CrossoverOperatorIndex];
if (crossOperatorType == GAParameters.CrossoverOperatorType.Arithmetic)
{
crossoverOperator = new ArithmeticCrossoverOperator(crossoverRate);
}
else
{
crossoverOperator = new HeuristicCrossoverOperator(crossoverRate);
}
var mutationOperatorType = GAParameters.MutationOperator[parameters.MutationOperatorIndex];
if (mutationOperatorType == GAParameters.MutationOperatorType.FloatingPointBoundary)
{
mutationOperator = new FloatingPointBoundaryMutationOperator(mutationRate);
}
else if (mutationOperatorType == GAParameters.MutationOperatorType.Uniform)
{
mutationOperator = new UniformMutationOperator(mutationRate);
}
else
{
mutationOperator = new GaussianMutationOperator(mutationRate);
}
if (GAParameters.SelectionOperator[parameters.SelectionOperatorIndex] == GAParameters.SelectionOperatorType.Tournament)
{
selectionOperator = new TournamentSelectionOperator <FloatingPointChromosome>(tournamentSize, problem);
}
else
{
selectionOperator = new RouletteWheelSelectionOperator <FloatingPointChromosome>(selectivePressure, problem);
}
IGeneticAlgorithm <FloatingPointChromosome> ga;
if (variant == GeneticAlgorithmVariant.Generational)
{
ga = new GenerationalGeneticAlgorithm <FloatingPointChromosome>(
problem,
populationSize,
selectionOperator,
crossoverOperator,
mutationOperator,
new CompositeTerminationCondition(
new MaxEvaluationsTerminationCondition(childEvaluations),
new MaxGenerationsTerminationCondition(maxGenerations)
),
elitismRate
);
}
else
{
ga = new SteadyStateGeneticAlgorithm <FloatingPointChromosome>(
problem,
populationSize,
selectionOperator,
crossoverOperator,
mutationOperator,
new CompositeTerminationCondition(
new MaxEvaluationsTerminationCondition(childEvaluations),
new MaxGenerationsTerminationCondition(maxGenerations)
),
generationGap,
new WorstFitnessReplacementOperator <FloatingPointChromosome>(problem)
);
}
// new GeneticAlgorithmRunner<FloatingPointChromosome>(ga).Run();
ga.Run();
childGAFitnesses[i] = ga.BestIndividual.Fitness;
}
}
chromosome.Fitness = childGAFitnesses.Average();
}
public GAParametersChromosome CreateChromosome()
{
return(new GAParametersChromosome(this, GAParameters.CreateRandom()));
}