public void Run()
{
// var populations = new int[]{30,50,100,200,500,1000,3000};
// Population and tournament size box plot data made with following alogorithm
var populationSizes = new int[] { 30, 50, 100, 200, 500, 1000, 3000 };
for (int j = 0; j < 10; j++)
{
string result = "";
for (int i = 0; i < populationSizes.Length; i++)
{
var populationSize = populationSizes[i];
var problem = new FunctionMinimizationFloatingPointProblem(new F6(), 2, -50, 150);
// var populationSize = 500;
var tournamentSize = 3;
var selectionOperator = new TournamentSelectionOperator <FloatingPointChromosome>(tournamentSize, problem);
var crossoverOperator = new HeuristicCrossoverOperator(0.8);
var mutationOperator = new GaussianMutationOperator(0.05);
var elitismRate = 1.0 / populationSize;
var ga = new GeneticAlgorithmBuilder <FloatingPointChromosome>(
problem,
crossoverOperator,
mutationOperator
)
.SetGeneticAlgorithmVariant(GeneticAlgorithmVariant.Generational)
.SetPopulationSize(populationSize)
.SetElitismRate(1.0 / populationSize)
.Build();
// new GeneticAlgorithmRunner<FloatingPointChromosome>(ga).Run();
ga.Run();
result += ga.BestIndividual.Fitness;
if (i != populationSizes.Length - 1)
{
result += ",";
}
}
Console.WriteLine(result);
}
}
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();
}
