private void AdaptMutationProbability(GeneticSharp.Domain.GeneticAlgorithm geneticAlgorithm)
{
var lastGenerations = geneticAlgorithm.Population.Generations.Take(10).ToList();
var bestChromosome = geneticAlgorithm.BestChromosome;
if (lastGenerations.Count < 10)
{
return;
}
bool haveSameFitness = lastGenerations.All(
x => Math.Abs(x.BestChromosome.Fitness.Value - bestChromosome.Fitness.Value) < float.Epsilon);
// all chromosomes have same fitness => increase mutation rate
if (haveSameFitness)
{
geneticAlgorithm.MutationProbability *= 1.01f;
geneticAlgorithm.MutationProbability =
Math.Min(geneticAlgorithm.MutationProbability, Global.Config.MaximumMutationProbability);
}
// otherwise decrease it if they don't have same fitness
else
{
geneticAlgorithm.MutationProbability /= 1.15f;
geneticAlgorithm.MutationProbability =
Math.Max(geneticAlgorithm.MutationProbability, Global.Config.MinimumMutationProbability);
}
}
public void GenerationRan(GeneticSharp.Domain.GeneticAlgorithm ga)
{
var latestFitness = 0.0;
ga.GenerationRan += (sender, e) =>
{
var bestChromosome = ga.BestChromosome as BinaryChromosome;
var bestFitness = bestChromosome.Fitness.Value;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
var phenotype = bestChromosome.GetGenes();
Console.WriteLine("Поколiння №{0,2}. Кращий результат = {1}", ga.GenerationsNumber, bestFitness);
Console.Write("Вид хромосоми: ");
foreach (GeneticSharp.Domain.Chromosomes.Gene g in phenotype)
{
Console.Write(g.Value.ToString() + "");
}
Console.WriteLine();
}
};
}
private ScheduleChromosome RunOnce(LineSeries lineSeries = null)
{
var adamChromosome = new ScheduleChromosome(jobShop);
var population = new Population(Global.Config.MinPopulationSize, Global.Config.MaxPopulationSize, adamChromosome);
var fitness = new ScheduleFitness();
var selection = new NonDeterministicTournamentSelection(Global.Config.TournamentSelectionProbability);
var crossover = new SchedulesCrossover(new CycleCrossover());
var mutation = new ScheduleMutation(Global.Config.MutationPerGeneProbability, new ReverseSequenceMutation());
var geneticAlgorithm =
new GeneticSharp.Domain.GeneticAlgorithm(population, fitness, selection, crossover, mutation)
{
Termination = new GenerationNumberTermination(Global.Config.GenerationsCount),
MutationProbability = Global.Config.MutationProbability,
CrossoverProbability = Global.Config.CrossoverProbability,
OperatorsStrategy = new JobShopOperatorStrategy(),
Reinsertion = new JobShopReinsertion(
new EliteSelection(),
new Elitism(Global.Config.ElitismPercent),
fitness
)
};
var stopWatch = new Stopwatch();
if (adaptive)
{
geneticAlgorithm.GenerationRan += (sender, e) => AdaptMutationProbability(geneticAlgorithm);;
}
geneticAlgorithm.GenerationRan += (sender, e) =>
{
Print(geneticAlgorithm.Population, stopWatch.Elapsed);
};
// plot the model
if (PlotModel != null)
{
geneticAlgorithm.GenerationRan += (sender, e) =>
{
lineSeries.Points.Add(new DataPoint(geneticAlgorithm.Population.GenerationsNumber,
((ScheduleChromosome)geneticAlgorithm.Population.BestChromosome).ScheduleLength.Value));
};
}
geneticAlgorithm.GenerationRan += (o, e) => GenerationRan?.Invoke(geneticAlgorithm.Population.CurrentGeneration);
if (Global.Config.ThreadsCount > 1)
{
geneticAlgorithm.TaskExecutor = new ParallelTaskExecutor()
{
MinThreads = 1,
MaxThreads = Global.Config.ThreadsCount
};
}
stopWatch.Start();
geneticAlgorithm.Start();
stopWatch.Stop();
return((ScheduleChromosome)geneticAlgorithm.BestChromosome);
}
public void BuildGA()
{
_ga = new GeneticSharp.Domain.GeneticAlgorithm(
_population,
_fitness,
_selection,
_crossover,
_mutation);
_ga.Termination = _termination;
}
public GeneticSharp.Domain.GeneticAlgorithm GetGA()
{
//Сам генетический алгоритм
var ga = new GeneticSharp.Domain.GeneticAlgorithm(
_population,
_fitness,
_selection,
_crossover,
_mutation);
ga.Termination = _termination;
return(ga);
}
public void Init(int patternLength, List <string> choices, int populationSize, int maxGenerations,
Selection selection, Crossover crossover, Mutation mutation)
{
string selectionname = Enum.GetName(typeof(Selection), selection);
_selection = (ISelection)Activator.CreateInstance("GeneticSharp.Domain", "GeneticSharp.Domain.Selections." + selectionname).Unwrap(); //new EliteSelection();
_crossover = (ICrossover)Activator.CreateInstance("GeneticSharp.Domain", "GeneticSharp.Domain.Crossovers." + crossover).Unwrap(); //new TwoPointCrossover();
_mutation = (IMutation)Activator.CreateInstance("GeneticSharp.Domain", "GeneticSharp.Domain.Mutations." + mutation).Unwrap(); //new PartialShuffleMutation();
_fittness = new ClosestToMillion();
_playPatternChromosome = new PlayPatternChromosome(patternLength, choices);
//var bestPattern = BestPattern.Get(context);
_population = new Population(populationSize, populationSize, _playPatternChromosome);
_geneticAlgorithm = new GeneticSharp.Domain.GeneticAlgorithm(_population, _fittness, _selection,
_crossover, _mutation);
_geneticAlgorithm.Termination = new GenerationNumberTermination(maxGenerations);
_geneticAlgorithm.Init();
}
public override void ConfigGA(GeneticSharp.Domain.GeneticAlgorithm ga)
{
var latestFitness = 0.0;
Context.GA.GenerationRan += (object sender, EventArgs e) =>
{
m_bestChromosome = Context.GA.BestChromosome as FloatingPointChromosome;
if (m_bestChromosome != null)
{
lock (m_positions)
{
var fitness = m_bestChromosome.Fitness.Value;
var points = m_bestChromosome.ToFloatingPoints();
m_positions.Add(new KeyValuePair <double, double[]>(fitness, points));
if (fitness != latestFitness)
{
latestFitness = fitness;
var phenotype = m_bestChromosome.ToFloatingPoints();
Console.WriteLine(
"Generation {0,2}: ({1},{2}),({3},{4}) = {5}",
ga.GenerationsNumber,
phenotype[0],
phenotype[1],
phenotype[2],
phenotype[3],
fitness
);
}
}
}
};
}
//Одне скомбіноване рішення
//Недоречний варіант!!!
public void Start()
{
//Сам генетический алгоритм
var ga = new GeneticSharp.Domain.GeneticAlgorithm(
_population,
_fitness,
_selection,
_crossover,
_mutation);
ga.Termination = _termination;
var latestFitness = 0.0;
ga.GenerationRan += (sender, e) =>
{
var bestChromosome = ga.BestChromosome as BinaryChromosome;
var bestFitness = bestChromosome.Fitness.Value;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
var phenotype = bestChromosome.GetGenes();
Console.WriteLine("Поколiння №{0,2}. Кращий результат = {1}", ga.GenerationsNumber, bestFitness);
Console.Write("Вид хромосоми: ");
foreach (GeneticSharp.Domain.Chromosomes.Gene g in phenotype)
{
Console.Write(g.Value.ToString() + "");
}
Console.WriteLine();
}
};
ga.Start();
}
static void Main(string[] args)
{
var selection = new EliteSelection();
var crossover = new OnePointCrossover();
var mutation = new ReverseSequenceMutation();
var fitness = new MyProblemFitness();
var chromosome = new MyProblemChromosome();
var population = new Population(20, 20, chromosome);
var ga = new GeneticSharp.Domain.GeneticAlgorithm(population, fitness, selection, crossover, mutation)
{
Termination = new GenerationNumberTermination(100)
};
Console.WriteLine("GA running...");
ga.Start();
if (ga.BestChromosome.Fitness != null)
{
double fitnessValue = ga.BestChromosome.Fitness.Value;
string solution = ga.BestChromosome.ToUserFriendly();
Console.WriteLine($"Best solution {solution} found has {fitnessValue} fitness for {ga.GenerationsNumber} generations.");
}
}
public void Start(GeneticSharp.Domain.GeneticAlgorithm ga) => ga.Start();
static void Main(string[] args)
{
TestData.GetData(out var servers, out var services);
var chromosome = new MyChromosome(servers, services);
var populationSize = 9;
var population = new Population(populationSize, populationSize, chromosome);
var fitness = new FuncFitness((c) =>
{
var mc = (MyChromosome)c;
var genes = mc.GetGenes();
var servers = genes.Select(x => (Server)x.Value).ToList();
var freeServersTerm = servers.Count(x => x.IsFree) * 2.5;
var negativeHddFreeTerm = servers.Where(x => x.HddFree < 0).Select(x => x.HddFree * 3).Sum();
var positiveHddFreeTerm = servers.Where(x => x.HddFree > 0).Select(x => x.HddFree * 0.2).Sum();
var negativeRamFreeTerm = servers.Where(x => x.RamFree < 0).Select(x => x.RamFree * 3).Sum();
var positiveRamFreeTerm = servers.Where(x => x.RamFree > 0).Select(x => x.RamFree * 0.2).Sum();
var negativeCpuFreeTerm = servers.Where(x => x.CpuFree < 0).Select(x => x.CpuFree * 3).Sum();
var positiveCpuFreeTerm = servers.Where(x => x.CpuFree > 0).Select(x => x.CpuFree * 0.2).Sum();
var fitness = freeServersTerm + negativeHddFreeTerm + positiveHddFreeTerm + negativeRamFreeTerm +
positiveRamFreeTerm + negativeCpuFreeTerm + positiveCpuFreeTerm;
//по каждому серверу нужно вычислить коэффициент заполненности
var serversFill = new List <double>();
foreach (var server in servers)
{
var serverFill = ((server.HddFull - server.HddFree) / server.HddFull) * 0.2 +
((server.RamFull - server.RamFree) / server.RamFull) * 0.4 +
((server.CpuFull - server.CpuFree) / server.CpuFull) * 0.4;
serversFill.Add(serverFill);
}
serversFill = serversFill.OrderByDescending(x => x).ToList();
//отладочная строка
//var serversFill = new List<double> { 0.9d, 0.9d, 0.6d, 0.2d,0 };
//среднее значение заполненности серверов
var averageFill = serversFill.Sum() / servers.Count();
//рассчитаем дисперсию у serversFill
var dispersion = serversFill.Select(eachFill => (eachFill - averageFill) * (eachFill - averageFill)).Sum() / servers.Count;
return((dispersion * 3) + fitness);
});
var selection = new EliteSelection();
//сделал свой кроссовер, похожий на UniformCrossover
var crossover = new MyCrossover();
//строка для отладки кроссовера
//crossover.Cross(new List<IChromosome> {chromosome.CreateNew(), chromosome.CreateNew()});
var mutation = new MyMutation();
//mutation.Mutate(chromosome, Single.Epsilon);
var termination = new FitnessStagnationTermination(100);
var ga = new GeneticSharp.Domain.GeneticAlgorithm(
population,
fitness,
selection,
crossover,
mutation)
{
Termination = termination,
MutationProbability = 0.3f
};
var latestFitness = 0.0;
ga.GenerationRan += (sender, e) =>
{
var ga = (GeneticSharp.Domain.GeneticAlgorithm)sender;
Console.Write("Номер итерации: " + ga.GenerationsNumber + " " + "Время работы: " + ga.TimeEvolving);
Console.WriteLine();
var bestChromosome = ga.BestChromosome as MyChromosome;
var bestFitness = bestChromosome.Fitness.Value;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
var phenotype = bestChromosome.GetGenes()
.Select(x => (Server)x.Value).ToList();
Console.Write("Фитнес: " + bestFitness);
Console.WriteLine();
foreach (var server in phenotype)
{
Console.Write("Сервер: " + server.Name +
", HddFull: " + server.HddFull +
", RamFull: " + server.RamFull +
", HddFree: " + server.HddFree +
", RamFree: " + server.RamFree +
", CpuFree: " + server.CpuFree +
", сервисы: ");
Console.Write(" Заполненность Hdd: " + Math.Round(100 - (server.HddFree / server.HddFull * 100)) + "%");
Console.Write(" Заполненность Ram: " + Math.Round(100 - (server.RamFree / server.RamFull * 100)) + "%");
Console.Write(" Заполненность Cpu: " + Math.Round(100 - server.CpuFree) + "%");
foreach (var service in server.Services)
{
Console.Write(service.Name + ", ");
}
Console.WriteLine();
Console.WriteLine();
//получить хромосомы на этой итерации и смотреть, равны ли они.
//Хромосомы равны, если равны их фитнесы.
//Если равны, алгоритм преждевременно сходится (вырождается популяция)
var chromosomesFitnessCollection = ga.Population.CurrentGeneration.Chromosomes;
}
}
};
ga.Start();
}
