protected ParameterOptimizationProblem(IParameterVectorEvaluator evaluator)
: base(evaluator, new UniformRandomRealVectorCreator())
{
Parameters.Add(new FixedValueParameter <IntValue>(ProblemSizeParameterName, "The dimension of the parameter vector that is to be optimized.", new IntValue(1)));
Parameters.Add(new ValueParameter <DoubleMatrix>(BoundsParameterName, "The bounds for each dimension of the parameter vector. If the number of bounds is smaller than the problem size then the bounds are reused in a cyclic manner.", new DoubleMatrix(new double[, ] {
{ 0, 100 }
}, new string[] { "LowerBound", "UpperBound" })));
Parameters.Add(new ValueParameter <StringArray>(ParameterNamesParameterName, "The element names which are used to calculate the quality of a parameter vector.", new StringArray(new string[] { "Parameter0" })));
SolutionCreator.LengthParameter.ActualName = "ProblemSize";
Operators.AddRange(ApplicationManager.Manager.GetInstances <IRealVectorOperator>());
strategyVectorCreator = new StdDevStrategyVectorCreator();
strategyVectorCreator.LengthParameter.ActualName = ProblemSizeParameter.Name;
strategyVectorCrossover = new StdDevStrategyVectorCrossover();
strategyVectorManipulator = new StdDevStrategyVectorManipulator();
strategyVectorManipulator.LearningRateParameter.Value = new DoubleValue(0.5);
strategyVectorManipulator.GeneralLearningRateParameter.Value = new DoubleValue(0.5);
Operators.Add(strategyVectorCreator);
Operators.Add(strategyVectorCrossover);
Operators.Add(strategyVectorManipulator);
Operators.Add(new BestSolutionAnalyzer());
Operators.Add(new BestSolutionsAnalyzer());
UpdateParameters();
UpdateStrategyVectorBounds();
RegisterEventHandlers();
}
public SingleObjectiveTestFunctionProblem()
: base(new AckleyEvaluator(), new UniformRandomRealVectorCreator())
{
Parameters.Add(new ValueParameter <DoubleMatrix>("Bounds", "The lower and upper bounds in each dimension.", Evaluator.Bounds));
Parameters.Add(new ValueParameter <IntValue>("ProblemSize", "The dimension of the problem.", new IntValue(2)));
Parameters.Add(new OptionalValueParameter <RealVector>("BestKnownSolution", "The best known solution for this test function instance."));
Maximization.Value = Evaluator.Maximization;
BestKnownQuality = new DoubleValue(Evaluator.BestKnownQuality);
strategyVectorCreator = new StdDevStrategyVectorCreator();
strategyVectorCreator.LengthParameter.ActualName = ProblemSizeParameter.Name;
strategyVectorCrossover = new StdDevStrategyVectorCrossover();
strategyVectorManipulator = new StdDevStrategyVectorManipulator();
strategyVectorManipulator.LearningRateParameter.Value = new DoubleValue(0.5);
strategyVectorManipulator.GeneralLearningRateParameter.Value = new DoubleValue(0.5);
SolutionCreator.RealVectorParameter.ActualName = "Point";
ParameterizeSolutionCreator();
ParameterizeEvaluator();
InitializeOperators();
RegisterEventHandlers();
UpdateStrategyVectorBounds();
}
protected ParameterOptimizationProblem(ParameterOptimizationProblem original, Cloner cloner)
: base(original, cloner)
{
strategyVectorCreator = cloner.Clone(original.strategyVectorCreator);
strategyVectorCrossover = cloner.Clone(original.strategyVectorCrossover);
strategyVectorManipulator = cloner.Clone(original.strategyVectorManipulator);
RegisterEventHandlers();
}
private SingleObjectiveTestFunctionProblem(SingleObjectiveTestFunctionProblem original, Cloner cloner)
: base(original, cloner)
{
strategyVectorCreator = cloner.Clone(original.strategyVectorCreator);
strategyVectorCrossover = cloner.Clone(original.strategyVectorCrossover);
strategyVectorManipulator = cloner.Clone(original.strategyVectorManipulator);
RegisterEventHandlers();
}
private EvolutionStrategy CreateEsGriewankSample()
{
EvolutionStrategy es = new EvolutionStrategy();
#region Problem Configuration
SingleObjectiveTestFunctionProblem problem = new SingleObjectiveTestFunctionProblem();
problem.ProblemSize.Value = 10;
problem.EvaluatorParameter.Value = new GriewankEvaluator();
problem.SolutionCreatorParameter.Value = new UniformRandomRealVectorCreator();
problem.Maximization.Value = false;
problem.Bounds = new DoubleMatrix(new double[, ] {
{ -600, 600 }
});
problem.BestKnownQuality.Value = 0;
problem.BestKnownSolutionParameter.Value = new RealVector(10);
problem.Name = "Single Objective Test Function";
problem.Description = "Test function with real valued inputs and a single objective.";
#endregion
#region Algorithm Configuration
es.Name = "Evolution Strategy - Griewank";
es.Description = "An evolution strategy which solves the 10-dimensional Griewank test function";
es.Problem = problem;
SamplesUtils.ConfigureEvolutionStrategyParameters <AverageCrossover, NormalAllPositionsManipulator,
StdDevStrategyVectorCreator, StdDevStrategyVectorCrossover, StdDevStrategyVectorManipulator>(
es, 20, 500, 2, 200, false);
StdDevStrategyVectorCreator strategyCreator = (StdDevStrategyVectorCreator)es.StrategyParameterCreator;
strategyCreator.BoundsParameter.Value = new DoubleMatrix(new double[, ] {
{ 1, 20 }
});
StdDevStrategyVectorManipulator strategyManipulator =
(StdDevStrategyVectorManipulator)es.StrategyParameterManipulator;
strategyManipulator.BoundsParameter.Value = new DoubleMatrix(new double[, ] {
{ 1E-12, 30 }
});
strategyManipulator.GeneralLearningRateParameter.Value = new DoubleValue(0.22360679774997896);
strategyManipulator.LearningRateParameter.Value = new DoubleValue(0.39763536438352531);
#endregion
return(es);
}
