public static void RunGeneticAlgorithm(List <Classifier> actionset, char[] condition, List <Classifier> population)
{
// Calculate sum over all Classifier in [A] with Timestamp * Numerosity
double sumTsNumerosity = 0;
double sumNumerosity = 0;
foreach (var classifier in actionset)
{
sumTsNumerosity += classifier.Ts * classifier.Numerosity;
sumNumerosity += classifier.Numerosity;
}
if (DateTime.Now.Ticks - sumTsNumerosity / sumNumerosity < XcsParamerter.ThetaGA)
{
// Update Timestamp of classifiers in [A]
foreach (var classifier in actionset)
{
classifier.Ts = DateTime.Now.Ticks;
}
Classifier parentOne = RouletteWheelSelection.SelectOffspring(actionset);
Classifier parentTwo = RouletteWheelSelection.SelectOffspring(actionset);
Classifier childOne = parentOne.Clone(parentOne);
Classifier childTwo = parentTwo.Clone(parentTwo);
childOne.Numerosity = childTwo.Numerosity = 1;
childOne.Experience = childTwo.Experience = 0;
// Do crossover
if (Helper.RandomGenerator.NextDouble() < XcsParamerter.Crossoverprobability)
{
OnePointCrossover.ApplyCrossover(childOne, childTwo);
childOne.Prediction = childTwo.Prediction = (parentOne.Prediction + parentTwo.Prediction) / 2;
childOne.Epsilon = childTwo.Epsilon = (parentOne.Epsilon + parentTwo.Epsilon) / 2;
childOne.Fitness = childTwo.Epsilon = (parentOne.Fitness + parentTwo.Fitness) / 2;
}
childOne.Fitness = childOne.Fitness * 0.1;
childTwo.Fitness = childTwo.Fitness * 0.1;
List <Classifier> children = new List <Classifier>()
{
childOne, childTwo
};
foreach (var child in children)
{
Mutation.ApplyMutation(child, condition);
if (XcsParamerter.DoGaSubsumption)
{
if (Subsumtion.DoesSubsume(parentOne, child))
{
parentOne.Numerosity++;
}
else if (Subsumtion.DoesSubsume(parentTwo, child))
{
parentTwo.Numerosity++;
}
else
{
InsertClassifierIntoPopulation(population, child);
}
}
else
{
InsertClassifierIntoPopulation(population, child);
}
DeleteFromPopulation(population);
}
}
}