public GeneSequence Clone()
{
var geneSequence = new GeneSequence(Genes.ToArray(), Strategy)
{
Fitness = Fitness
};
if (_stringGenes !=null)
{
geneSequence._stringGenes = _stringGenes;
}
return geneSequence;
}
private void PrintBest(int generation, GeneSequence geneSequence)
{
DisplayGenes(1 + generation, geneSequence.Fitness.Value, geneSequence.GetStringGenes(), geneSequence.Strategy.Description);
}
private static int FitnessSort(GeneSequence x, GeneSequence y)
{
int result = x.Fitness.Value.CompareTo(y.Fitness.Value);
if (result == 0)
{
result = y.Generation.CompareTo(x.Generation);
}
return result;
}
public GeneSequence GetBestGenetically(int numberOfGenesToUse, string possibleGenes, Func<string, FitnessResult> calcFitness)
{
int seed = _randomSeed != 0 ? _randomSeed : (int)DateTime.Now.Ticks;
Console.WriteLine("using random seed: " + seed);
_random = new Random(seed);
var popAlpha = new List<GeneSequence>(GaPopsize);
var popBeta = new List<GeneSequence>(GaPopsize);
Func<char> getRandomGene = () => possibleGenes[_random.Next(possibleGenes.Length)];
InitPopulation(UseHillClimbing ? NumberOfGenesInUnitOfMeaning : numberOfGenesToUse, popAlpha, popBeta, getRandomGene);
var population = popAlpha;
var spare = popBeta;
var previousBests = new List<GeneSequence>
{
population.First()
};
previousBests[0].Fitness = calcFitness(previousBests[0].GetStringGenes());
int generation = 0;
var generationsBetweenImprovments = Enumerable.Repeat(_gaMaxGenerationsWithoutImprovement, 20).ToList();
if (UseHillClimbing)
{
int failureToImproveCount = 0;
var bestEver = new GeneSequence(new char[] { }, null)
{
Fitness = new FitnessResult{Value = UInt32.MaxValue}
};
for (int i = NumberOfGenesInUnitOfMeaning; i < numberOfGenesToUse - 1; i += NumberOfGenesInUnitOfMeaning)
{
GetBestGenetically(OnlyPermuteNewGenesWhileHillClimbing ? i - NumberOfGenesInUnitOfMeaning : 0, i, getRandomGene, previousBests, population, spare, calcFitness, generationsBetweenImprovments, ref generation);
var incrementalBest = previousBests.First();
if (incrementalBest.Fitness.Value < bestEver.Fitness.Value)
{
bestEver = incrementalBest.Clone();
failureToImproveCount = 0;
}
else
{
failureToImproveCount++;
if (incrementalBest.Fitness.Value > bestEver.Fitness.Value * 1.05m || failureToImproveCount >= 5)
{
Console.WriteLine("Fitness appears to be getting worse, returning best result: fitness " + bestEver.Fitness.Value + " in generation " + bestEver.Generation);
return bestEver;
}
}
var parents = previousBests.ToList();
population.Clear();
spare.Clear();
previousBests.Clear();
for (int j = 0; j < GaPopsize; j++)
{
var random = _randomStrategy.Generate(null, NumberOfGenesInUnitOfMeaning, getRandomGene, NumberOfGenesInUnitOfMeaning, _slidingMutationRate, _random.Next, 0);
var newChild = new GeneSequence(parents[j % parents.Count].Genes.Concat(random.Genes).ToArray(), _mutationStrategy)
{
Fitness = GeneSequence.DefaultFitness
};
population.Add(newChild);
spare.Add(newChild.Clone());
}
previousBests.AddRange(population.Select(x =>
{
x.Fitness = calcFitness(x.GetStringGenes());
return x;
}).Where(x => x.Fitness.Value < incrementalBest.Fitness.Value));
if (previousBests.Count < 100)
{
previousBests.AddRange(population.OrderBy(x => x.Fitness.Value).Take(100 - previousBests.Count));
}
SortByFitness(previousBests);
int count = 1 + (i / NumberOfGenesInUnitOfMeaning);
Console.WriteLine("> " + count);
if (previousBests[0].Fitness.Value < incrementalBest.Fitness.Value)
{
PrintBest(generation, previousBests[0]);
}
}
return bestEver;
}
GetBestGenetically(0, numberOfGenesToUse, getRandomGene, previousBests, population, spare, calcFitness, generationsBetweenImprovments, ref generation);
var best = previousBests.First();
return best;
}