public void CanGenerateGenoType()
{
var randomGenerator = new UniformRandomGenerator();
var possibleFunctions = new List <IGenoTypeNode>
{
new SquareRoot(),
new Multiplication(),
new Plus(),
new Minus()
};
var possibleTerminals = new List <IGenoTypeNode>
{
new FeatureTerminal("a"),
new FeatureTerminal("b"),
new FeatureTerminal("c"),
new FeatureTerminal("d")
};
var eaGeneExpressionParameters = new EaGeneExpressionParameters(4, possibleFunctions, possibleTerminals);
var parameterTerminalFactory = new ParameterTerminalFactory(eaGeneExpressionParameters, randomGenerator);
var genoTypeFactory = new GenoTypeFactory(eaGeneExpressionParameters, randomGenerator, parameterTerminalFactory);
var genoType = genoTypeFactory.GetGenoType();
Assert.NotNull(genoType);
}
public void todo(IDataSet dataSet)
{
var randomGenerator = new UniformRandomGenerator();
var possibleFunctions = new List <IGenoTypeNode>
{
new Multiplication(),
new Plus(),
new Minus()
};
var possibleTerminals = new List <IGenoTypeNode>();
foreach (var mappedColumn in dataSet.MappedColumns)
{
possibleTerminals.Add(new FeatureTerminal(mappedColumn.Key));
}
var eaGeneExpressionParameters = new EaGeneExpressionParameters(10, possibleFunctions, possibleTerminals);
var parameterTerminalFactory = new ParameterTerminalFactory(eaGeneExpressionParameters, randomGenerator);
var genoTypeFactory = new GenoTypeFactory(eaGeneExpressionParameters, randomGenerator, parameterTerminalFactory);
var genoType = genoTypeFactory.GetGenoType();
var phenoTypeTree = new PhenoTypeTree(genoType.GenoTypeNodes);
var stringExpresssion = phenoTypeTree.ToString();
// TODO make mapped columns more sophisticated
var mappedColumnsUsedInExpression = new Dictionary <string, int>();
foreach (var mappedColumn in dataSet.MappedColumns)
{
if (stringExpresssion.Contains(mappedColumn.Key))
{
mappedColumnsUsedInExpression.Add(mappedColumn.Key, mappedColumn.Value);
}
}
var expression = new Expression(stringExpresssion);
var numberOfRows = dataSet.MappedData.GetLength(0);
var sum = 0.0;
for (var row = 0; row < numberOfRows; row++)
{
foreach (var usedMappedColumn in mappedColumnsUsedInExpression)
{
expression.Parameters[usedMappedColumn.Key.Replace("]", "").Replace("[", "")] = dataSet.MappedData[row, usedMappedColumn.Value];
}
if (!expression.HasErrors())
{
var test = (double)expression.Evaluate();
sum = sum + test;
}
}
}
public IGenoTypeNode GetTerminalNode()
{
if (UniformRandomGenerator.GetContinousRandomNumber(0, 1.0) <
EaGeneExpressionParameters.ParameterProbability)
{
return(ParameterTerminalFactory.GetParameterTerminal());
}
return(EaGeneExpressionParameters.PossibleTerminals[
UniformRandomGenerator.GetIntegerRandomNumber(0, NumberOfPossibleTerminals)]);
}
public void CanTranslateTest3()
{
var randomGenerator = new UniformRandomGenerator();
var possibleFunctions = new List <IGenoTypeNode>
{
new SquareRoot(),
new Multiplication(),
new Division(),
new Plus(),
new Minus(),
new Minimum(),
new Maximum(),
new Not(),
new Exp(),
new Sinus(),
new Cosinus()
};
var possibleTerminals = new List <IGenoTypeNode>
{
new FeatureTerminal("a"),
new FeatureTerminal("b"),
new FeatureTerminal("c"),
new FeatureTerminal("d")
};
var eaGeneExpressionParameters = new EaGeneExpressionParameters(20, possibleFunctions, possibleTerminals);
var parameterTerminalFactory = new ParameterTerminalFactory(eaGeneExpressionParameters, randomGenerator);
var genoTypeFactory = new GenoTypeFactory(eaGeneExpressionParameters, randomGenerator, parameterTerminalFactory);
eaGeneExpressionParameters.ParameterTypeInteger = true;
eaGeneExpressionParameters.ConstantProbability = 0;
var genoType = genoTypeFactory.GetGenoType();
var phenoTypeTree = new GeneExpression.PhenoTypeTree(genoType.GenoTypeNodes);
var expresssion = phenoTypeTree.ToString();
}
static void Main()
{
var randomGenerator = new UniformRandomGenerator();
var dataSet = GetDataSet();
var target = GetTarget();
var eaGeneExpressionParameters = GetEaGeneExpressionParameters(dataSet);
var parameterTerminalFactory = new ParameterTerminalFactory(eaGeneExpressionParameters, randomGenerator);
var genoTypeFactory = new GenoTypeFactory(eaGeneExpressionParameters, randomGenerator, parameterTerminalFactory);
var mutator = new GenoTypeMutatorBasic1(randomGenerator, eaGeneExpressionParameters, genoTypeFactory);
var crossOverator = new GenoTypeCrossoveratorBasic1(randomGenerator, eaGeneExpressionParameters);
var populationP = GetFirstPopulation(eaGeneExpressionParameters, parameterTerminalFactory, genoTypeFactory);
var listOfObjectiveValuesP = new List <IObjectiveValues>();
for (var c = 0; c < populationP.Count; c++)
{
var objectiveValues = GetObjectiveValues(target, dataSet, populationP[c]);
listOfObjectiveValuesP.Add(objectiveValues);
}
listOfObjectiveValuesP = Nsga2TournamentSelector.PerformSelection(eaGeneExpressionParameters.TournamentSize, listOfObjectiveValuesP, randomGenerator);
var tempPopulation = new List <Individual>();
foreach (var objectiveValues in listOfObjectiveValuesP)
{
var indy = populationP.FirstOrDefault(x => x.Guid == objectiveValues.IndividualGuid);
tempPopulation.Add((Individual)indy.Clone());
}
populationP = tempPopulation;
for (var generation = 0; generation < eaGeneExpressionParameters.NumberOfGeneration; generation++)
{
var populationQ = new List <Individual>();
foreach (var individual in populationP)
{
populationQ.Add((Individual)individual.Clone());
}
populationQ = PerformCrossOver(populationQ, randomGenerator, crossOverator);
foreach (var individual in populationQ)
{
mutator.PerformMutation(ref individual.GenoType);
}
var combinedPopulation = new List <Individual>();
var combinedlistOfObjectiveValues = new List <IObjectiveValues>();
foreach (var individual in populationP)
{
var clone = (Individual)individual.Clone();
combinedPopulation.Add(clone);
var objectiveValues = GetObjectiveValues(target, dataSet, clone);
combinedlistOfObjectiveValues.Add(objectiveValues);
}
foreach (var individual in populationQ)
{
var clone = (Individual)individual.Clone();
combinedPopulation.Add(clone);
var objectiveValues = GetObjectiveValues(target, dataSet, clone);
combinedlistOfObjectiveValues.Add(objectiveValues);
}
combinedlistOfObjectiveValues = Nsga2Ranker.Rank(combinedlistOfObjectiveValues);
combinedlistOfObjectiveValues = Nsga2Crowder.CalculateCrowdingDistances(combinedlistOfObjectiveValues);
combinedlistOfObjectiveValues = combinedlistOfObjectiveValues.OrderBy(i => i.Rank).ThenByDescending(i => i.CrowdingDistance).ToList();
tempPopulation = new List <Individual>();
var counter = 0;
var smallestMse = decimal.MaxValue;
decimal largestMse = 0;
var smallestPosition = -1;
foreach (var objectiveValues in combinedlistOfObjectiveValues)
{
var indy = combinedPopulation.FirstOrDefault(x => x.Guid == objectiveValues.IndividualGuid);
tempPopulation.Add((Individual)indy.Clone());
var ovs = GetObjectiveValues(target, dataSet, combinedPopulation.FirstOrDefault(x => x.Guid == objectiveValues.IndividualGuid));
if (ovs.Values[0] < smallestMse)
{
smallestMse = ovs.Values[0];
smallestPosition = counter;
}
if (ovs.Values[0] > largestMse)
{
largestMse = ovs.Values[0];
}
counter++;
if (counter == eaGeneExpressionParameters.PopulationSize)
{
break;
}
}
populationP = tempPopulation;
Console.WriteLine(generation + " " + smallestMse + " " + largestMse);
Console.WriteLine(new PhenoTypeTree(populationP[smallestPosition].GenoType.GenoTypeNodes));
}
}
