public static Genome MutateGenome(Genome genome, double mutationRate)
{
for (int i = 0; i < genome.Phrase.Length; i++)
{
genome.Phrase[i] = Random.NextDouble() < mutationRate
? RandomGenomeGenerator.RandomChars(1).First()
: genome.Phrase[i];
}
return genome;
}
public static int GetFitness(Genome genome, string answer)
{
var fitness = answer.Where((currentChar, i) => currentChar == genome[i]).Count();
genome.Fitness = fitness;
return fitness;
}
public static Genome CrossoverGenomes(Genome genome1, Genome genome2)
{
return new Genome(genome1.Phrase.Zip(genome2.Phrase, (g1, g2) => Random.Next(0,2) == 0 ? g1 : g2).ToArray());
}
public Genome <GType, PType> Crossover(Genome <GType, PType> partner)
{
return(InnerCrossover(partner));
}
protected abstract Genome <GType, PType> InnerCrossover(Genome <GType, PType> partner);
public void NextGeneration(double rangeRatio, out double newRangeRatio)
{
#region Count the number of generations where the result is stuck at some local optimum.
if (Generation == 1)
{
stuckFitnessValue = maxFitness;
stuckGenerationNumberLimit = initialStuckGenerationNumberLimit;
rangeRatioIncrement = initialRangeRatioIncrement;
}
else
{
if (Math.Round(stuckFitnessValue, 6) != Math.Round(maxFitness, 6))//Reduced the precision of fiteness value comparison. He, 16/07/2010.
{
stuckFitnessValue = maxFitness;
stuckGenerationNumber = 0;
stuckGenerationNumberLimit = initialStuckGenerationNumberLimit;
rangeRatioIncrement = initialRangeRatioIncrement;
}
else
{
stuckGenerationNumber++;
}
}
#endregion
#region Calculate the current range ratio of the parameters.
double deltaRangeRatio = Convert.ToDouble((double)(1) / (double)numberOfRangeSteps);
deltaRangeRatio = deltaRangeRatio * rangeShrinkingFactor;
rangeRatio = Math.Max((rangeRatio - deltaRangeRatio), rangeRatioLimit);//Don't let the range less than 10% of the initial range.
if (stuckGenerationNumber == stuckGenerationNumberLimit)
{
rangeRatio = rangeRatio + rangeRatioIncrement;
stuckGenerationNumber = 0;
stuckGenerationNumberLimit = stuckGenerationNumberLimit + stuckGenerationNumberLimitIncrement;
rangeRatioIncrement = rangeRatioIncrement + deltaRangeRatioIncrement;
}
newRangeRatio = rangeRatio;
Console.WriteLine("------------------------");
Console.WriteLine(stuckGenerationNumber);
Console.WriteLine(stuckGenerationNumberLimit);
Console.WriteLine(rangeRatio);
Console.WriteLine("------------------------");
#endregion
// increment the generation;
Generation++;
// check who can die
for (int i = 0; i < Genomes.Count; i++)
{
if (((Genome)Genomes[i]).CanDie(kDeathFitness))
{
Genomes.RemoveAt(i);
i--;
}
}
// determine who can reproduce
GenomeReproducers.Clear();
GenomeResults.Clear();
for (int i = 0; i < Genomes.Count; i++)
{
if (((Genome)Genomes[i]).CanReproduce(kReproductionFitness))
{
GenomeReproducers.Add(Genomes[i]);
}
}
ArrayList tempGenomes = (ArrayList)GenomeReproducers.Clone();
tempGenomes.Sort(myCompareMethod);
tempGenome0 = (Genome)tempGenomes[0];
parameterValueArray0.Clear();
parameterValueArray0 = getParameterValues(tempGenome0); //Save the genome that has the highest fitness value in this generation,which is the first one after sorting.
int crossoverIncrementRatio = Math.Min((stuckGenerationNumberLimit / initialStuckGenerationNumberLimit), (populationIncrementRatioLimit + 1));
for (int i = 0; i < crossoverIncrementRatio; i++)
{
DoCrossover(GenomeReproducers);
}
Genomes = (ArrayList)GenomeResults.Clone();
Console.WriteLine("================================================================");
Console.WriteLine("The number of genomes after crossover is " + Genomes.Count);
Console.WriteLine("================================================================");
// mutate a part of genes, which is determined by the frequency, in the new population, except the last two genes which are the best two from laste generation.
for (int i = 0; i < (Genomes.Count); i++)
{
Mutate(
(Genome)Genomes[i],
kParameterMin,
kParameterMax,
kParameterStep,
kParameterStepLimit,
Generation,
optimalParameterValueArray,
rangeRatio);
}
//Generate local genomes, i.e. local exploitation of best solution.
if (maxFitness > fitnessShresholdForLoacalSearch) //Only start local search when the maxFitnees is greater than the threshold value.
{
LocalPopulationGeneration();
}
Genomes.Add(tempGenome0);
setParmeterValues((Genome)Genomes[Genomes.Count - 1], parameterValueArray0);
//Add the best genome from previous generations...
//calculate fitness of all the genes
for (int i = 0; i < (Genomes.Count); i++)
{
double CurrentFitness = ((Genome)Genomes[i]).CalculateFitness(calibrationRoundID, dataDirectory, workDirectory, projectDirectory, genericAlgorithmsDirectory, observationArray, stdArray, kParameterSiteIndex, amplificationFactor, CV, observationDateArray, observationSeperator);
}
Genomes.Sort(myCompareMethod);
// kill all the genes above the population limit
if (Genomes.Count > kPopulationLimit)
{
Genomes.RemoveRange(kPopulationLimit, Genomes.Count - kPopulationLimit);
}
CurrentPopulation = Genomes.Count;
//Console.WriteLine(CurrentPopulation);
}
private void setParmeterValues(Genome aGene, ArrayList parameterVaueList)
{
aGene.setParameterValues(parameterVaueList);
}
public static Genome CrossoverGenomes(Genome genome1, Genome genome2)
{
return(new Genome(genome1.Phrase.Zip(genome2.Phrase, (g1, g2) => Random.Next(0, 2) == 0 ? g1 : g2).ToArray()));
}
