public void Start_AnyTaskWithException_Exception()
{
var pipeline = "";
var target = new ParallelTaskExecutor();
target.Add(() =>
{
throw new Exception("1");
});
target.Add(() =>
{
Thread.Sleep(5);
pipeline += "2";
});
target.Add(() =>
{
Thread.Sleep(5);
pipeline += "3";
});
Assert.Catch <Exception>(() =>
{
target.Start();
}, "1");
}
public void Start_ParallelManySlowFitness_Timeout()
{
var taskExecutor = new ParallelTaskExecutor();
taskExecutor.MinThreads = 100;
taskExecutor.MaxThreads = 100;
taskExecutor.Timeout = TimeSpan.FromMilliseconds(1000);
var selection = new RouletteWheelSelection();
var crossover = new OnePointCrossover(1);
var mutation = new UniformMutation();
var chromosome = new ChromosomeStub();
var target = new GeneticAlgorithm(new Population(100, 150, chromosome),
new FitnessStub()
{
SupportsParallel = true, ParallelSleep = 1500
}, selection, crossover, mutation);
target.TaskExecutor = taskExecutor;
Assert.Catch <TimeoutException>(() =>
{
target.Start();
}, "The fitness evaluation reached the 00:00:01 timeout.");
Assert.IsFalse(target.IsRunning);
Assert.AreEqual(GeneticAlgorithmState.Stopped, target.State);
}
public void Start_ManyTasksWithGreaterNumberOfThreads_ParallelExecuted()
{
var pipeline = "";
var target = new ParallelTaskExecutor {
MinThreads = int.MaxValue, MaxThreads = int.MaxValue
};
target.Add(() =>
{
pipeline += "1";
});
target.Add(() =>
{
Thread.Sleep(100);
pipeline += "2";
});
target.Add(() =>
{
Thread.Sleep(10);
pipeline += "3";
});
var actual = target.Start();
Assert.IsTrue(actual);
Assert.AreEqual("132", pipeline);
}
public void Stop_Tasks_ShutdownCalled()
{
var pipeline = "";
var target = new ParallelTaskExecutor
{
Timeout = TimeSpan.FromMilliseconds(1000)
};
target.Add(() =>
{
Thread.Sleep(500);
pipeline += "1";
});
target.Add(() =>
{
Thread.Sleep(500);
pipeline += "2";
});
target.Add(() =>
{
Thread.Sleep(500);
pipeline += "3";
});
Parallel.Invoke(
() => target.Start(),
() =>
{
Thread.Sleep(100);
target.Stop();
});
Assert.IsFalse(target.IsRunning);
}
public void Stop_ManyTasks_StopAll()
{
var pipeline = "";
var target = new ParallelTaskExecutor();
target.Timeout = TimeSpan.FromMilliseconds(1000);
target.Add(() =>
{
Thread.Sleep(5);
pipeline += "1";
});
target.Add(() =>
{
Thread.Sleep(5);
pipeline += "2";
});
target.Add(() =>
{
Thread.Sleep(5);
pipeline += "3";
});
Parallel.Invoke(
() => Assert.IsTrue(target.Start()),
() =>
{
Thread.Sleep(100);
target.Stop();
});
}
public void Start_ManyTasks_ParallelExecuted()
{
var pipeline = "";
var target = new ParallelTaskExecutor();
target.Add(() =>
{
pipeline += "1";
});
target.Add(() =>
{
Thread.Sleep(100);
pipeline += "2";
});
target.Add(() =>
{
Thread.Sleep(10);
pipeline += "3";
});
var actual = target.Start();
Assert.IsTrue(actual);
Assert.AreEqual("132", pipeline);
}
public void Initialize(IOptimizerConfiguration config, OptimizerFitness fitness)
{
_config = config;
_fitness = fitness;
_executor = new ParallelTaskExecutor()
{
MinThreads = 1
};
_executor.MaxThreads = _config.MaxThreads > 0 ? _config.MaxThreads : 8;
}
public void Start_ParallelManyGenerations_Optimization()
{
var taskExecutor = new ParallelTaskExecutor();
taskExecutor.MinThreads = 100;
taskExecutor.MaxThreads = 100;
var selection = new EliteSelection();
var crossover = new OnePointCrossover(1);
var mutation = new UniformMutation();
var chromosome = new ChromosomeStub();
FlowAssert.IsAtLeastOneAttemptOk(20, () =>
{
var target = new GeneticAlgorithm(new Population(100, 150, chromosome),
new FitnessStub()
{
SupportsParallel = true
}, selection, crossover, mutation);
target.TaskExecutor = taskExecutor;
Assert.AreEqual(GeneticAlgorithmState.NotStarted, target.State);
Assert.IsFalse(target.IsRunning);
target.Start();
Assert.AreEqual(GeneticAlgorithmState.TerminationReached, target.State);
Assert.IsFalse(target.IsRunning);
Assert.IsTrue(target.Population.CurrentGeneration.Chromosomes.Count >= 100);
Assert.IsTrue(target.Population.CurrentGeneration.Chromosomes.Count <= 150);
Assert.IsNotNull(target.Population.BestChromosome);
Assert.IsTrue(target.Population.BestChromosome.Fitness >= 0.9);
Assert.IsTrue(target.Population.Generations.Count > 0);
});
FlowAssert.IsAtLeastOneAttemptOk(20, () =>
{
var target = new GeneticAlgorithm(new Population(100, 150, chromosome),
new FitnessStub()
{
SupportsParallel = true
}, selection, crossover, mutation);
target.TaskExecutor = taskExecutor;
target.Start();
Assert.IsTrue(target.Population.CurrentGeneration.Chromosomes.Count >= 100);
Assert.IsTrue(target.Population.CurrentGeneration.Chromosomes.Count <= 150);
Assert.IsNotNull(target.Population.BestChromosome);
Assert.IsTrue(target.Population.BestChromosome.Fitness >= 0.9);
Assert.IsTrue(target.Population.Generations.Count > 0);
});
}
public void Start_Timeout_False()
{
var pipeline = "1";
var target = new ParallelTaskExecutor();
target.Timeout = TimeSpan.FromMilliseconds(2);
target.Add(() =>
{
Thread.Sleep(10);
pipeline += "2";
});
var actual = target.Start();
Assert.IsFalse(actual);
}
public void Start_MaxThreads1_DoNotBlockOtherThreads()
{
var target = new ParallelTaskExecutor
{
MinThreads = 1,
MaxThreads = 1
};
target.Add(() =>
{
});
target.Add(() =>
{
Thread.Sleep(200);
});
target.Add(() =>
{
Thread.Sleep(10);
});
int otherThreadCount = 0;
var otherThread = new System.Timers.Timer(50)
{
AutoReset = true
};
otherThread.Elapsed += (sender, arg) =>
{
otherThreadCount++;
};
otherThread.Start();
Task.Run(() =>
{
target.Start();
}).Wait();
otherThread.Stop();
Assert.GreaterOrEqual(otherThreadCount, 2);
}
public static FloatingPointChromosome EvolveGeneticAlgorithm(FloatingPointChromosome chromosome, IOthelloAgent agent, string chromosomeLabel = "")
{
IPopulation population = new TplPopulation(30, 60, chromosome);
IFitness fitness = new EvaluationFitness(agent);
ISelection selection = new RouletteWheelSelection(); //Guess
ICrossover crossover = new UniformCrossover(); //Guess
IMutation mutation = new UniformMutation(); //Guess
ITermination stagnation = new FitnessStagnationTermination(500);
ITermination threshold = new FitnessThresholdTermination(.9);
ITaskExecutor taskExecutor = new ParallelTaskExecutor()
{
MaxThreads = Environment.ProcessorCount,
MinThreads = Environment.ProcessorCount / 2
};
GeneticAlgorithm algorithm = new GeneticAlgorithm(population, fitness, selection, crossover, mutation)
{
TaskExecutor = new TplTaskExecutor(),
MutationProbability = .2f
};
algorithm.TaskExecutor = taskExecutor;
algorithm.Termination = stagnation;
algorithm.Start();
SaveChromosome((FloatingPointChromosome)algorithm.BestChromosome, chromosomeLabel);
Debug.WriteLine("finished Training, with {0} time spent on evolving", algorithm.TimeEvolving);
Debug.WriteLine("fitness of this generation vs the last : {0}", algorithm.Fitness);
return((FloatingPointChromosome)algorithm.BestChromosome);
}
private void EvolveGeneticStrategyButton_Click(object sender, RoutedEventArgs e)
{
OutputTextBlock.Text = "Evolving...";
Task.Run(() =>
{
var chromosome = new BlackjackChromosome();
var fitness = new BlackjackFitness();
var population = new Population(Settings.GeneticSettings.MinPopulationSize, Settings.GeneticSettings.MaxPopulationSize, chromosome);
ISelection selection;
switch (Settings.GeneticSettings.SelectionType)
{
case SelectionType.Elite:
selection = new EliteSelection();
break;
case SelectionType.RouletteWheel:
selection = new RouletteWheelSelection();
break;
case SelectionType.StochasticUniversalSampling:
selection = new StochasticUniversalSamplingSelection();
break;
case SelectionType.Tournament:
selection = new TournamentSelection(Settings.GeneticSettings.TournamentSize);
break;
default:
throw new InvalidOperationException();
}
ICrossover crossover;
switch (Settings.GeneticSettings.CrossoverType)
{
case CrossoverType.AlternatingPosition:
crossover = new AlternatingPositionCrossover();
break;
case CrossoverType.CutAndSplice:
crossover = new CutAndSpliceCrossover();
break;
case CrossoverType.Cycle:
crossover = new CycleCrossover();
break;
case CrossoverType.OnePoint:
crossover = new OnePointCrossover();
break;
case CrossoverType.TwoPoint:
crossover = new TwoPointCrossover();
break;
case CrossoverType.OrderBased:
crossover = new OrderBasedCrossover();
break;
case CrossoverType.Ordered:
crossover = new OrderedCrossover();
break;
case CrossoverType.PartiallyMapped:
crossover = new PartiallyMappedCrossover();
break;
case CrossoverType.PositionBased:
crossover = new PositionBasedCrossover();
break;
case CrossoverType.ThreeParent:
crossover = new ThreeParentCrossover();
break;
case CrossoverType.Uniform:
crossover = new UniformCrossover(Settings.Current.GeneticSettings.MixProbability);
break;
case CrossoverType.VotingRecombination:
crossover = new VotingRecombinationCrossover();
break;
default:
throw new InvalidOperationException();
}
var mutation = new UniformMutation();
var termination = new FitnessStagnationTermination(Settings.Current.GeneticSettings.NumStagnantGenerations);
var taskExecutor = new ParallelTaskExecutor();
var ga = new GeneticAlgorithm(
population,
fitness,
selection,
crossover,
mutation);
ga.Termination = termination;
ga.TaskExecutor = taskExecutor;
ga.MutationProbability = Settings.GeneticSettings.MutationProbability;
ga.CrossoverProbability = Settings.GeneticSettings.CrossoverProbability;
var latestFitness = double.MinValue;
ga.GenerationRan += (s, o) =>
{
geneticStrategy = (IStrategy)ga.BestChromosome;
var generationNumber = ga.GenerationsNumber;
var bestFitness = ga.BestChromosome.Fitness.Value;
var avgFitness = ga.Population.CurrentGeneration.Chromosomes.Average(c => c.Fitness.Value);
Dispatcher.Invoke(() =>
{
if (generationNumber == 1)
{
OutputTextBlock.Text = string.Empty;
}
OutputTextBlock.Text = $"Gen: {generationNumber}\tFit: {bestFitness}\tAvg: {avgFitness.ToString("0")}\n" + OutputTextBlock.Text;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
var savedImageName = Settings.Current.GeneticSettings.SaveImagePerGeneration ? "gen" + generationNumber : null;
StrategyViewer.Draw(GeneticStrategyCanvas, geneticStrategy, $"Best from generation {generationNumber}", savedImageName);
}
}, DispatcherPriority.Background);
};
ga.TerminationReached += (s, o) =>
{
Dispatcher.Invoke(() =>
{
OutputTextBlock.Text = "Termination reached.\n" + OutputTextBlock.Text;
TestGeneticStrategyButton.IsEnabled = true;
}, DispatcherPriority.Background);
};
ga.Start();
});
}
public void Optimize(bool parallel = false)
{
Console.WriteLine("Running optimization");
// Rank StopGain StopLoss MinEpsGrowth
var chromosome = new SimChromosome();
var population = new Population(25, 50, chromosome);
var fitness = new SimFitness(List, Dbfile);
var selection = new EliteSelection();
var crossover = new UniformCrossover(0.5f);
var mutation = new UniformMutation();
var termination = new FitnessStagnationTermination(100);
var ga = new GeneticAlgorithm(
population,
fitness,
selection,
crossover,
mutation);
if (parallel)
{
var taskExecutor = new ParallelTaskExecutor();
taskExecutor.MinThreads = 4;
taskExecutor.MaxThreads = 12;
ga.TaskExecutor = taskExecutor;
}
ga.Termination = termination;
Console.WriteLine("Generation: (minRank, stopGain, stopLoss) = value");
var latestFitness = 0.0;
ga.GenerationRan += (sender, e) =>
{
var bestChromosome = ga.BestChromosome as SimChromosome;
var bestFitness = bestChromosome.Fitness.Value;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
var Rank = (double)bestChromosome.GetGene(0).Value;
var StopGain = (double)bestChromosome.GetGene(1).Value;
var StopLoss = (double)bestChromosome.GetGene(2).Value;
var MinGrowth = (double)bestChromosome.GetGene(3).Value;
Console.WriteLine(
"Generation {0,2}: ({1},{2},{3}) = {4}",
ga.GenerationsNumber,
Rank,
StopGain,
StopLoss,
bestFitness
);
}
};
ga.Start();
}
/// <summary>
/// Initializes a new instance of the <see cref="AlgorithmOptimumFinder"/> class
/// </summary>
/// <param name="start">Algorithm start date</param>
/// <param name="end">Algorithm end date</param>
/// <param name="fitScore">Argument of <see cref="FitnessScore"/> type. Fintess function to rank the backtest results</param>
/// <param name="filterEnabled">Indicates whether to apply fitness filter to backtest results</param>
public AlgorithmOptimumFinder(DateTime start, DateTime end, FitnessScore fitScore, bool filterEnabled)
{
// Assign Dates and Criteria to sort the results
StartDate = start;
EndDate = end;
FitnessScore = fitScore;
// Common properties
var selection = new RouletteWheelSelection();
// Properties specific to optimization modes
IFitness fitness;
PopulationBase population;
ITaskExecutor executor;
ITermination termination;
// Task execution mode
switch (Shared.Config.TaskExecutionMode)
{
// Enable fitness filtering while searching for optimum parameters
case TaskExecutionMode.Linear:
executor = new LinearTaskExecutor();
fitness = new OptimizerFitness(StartDate, EndDate, fitScore, filterEnabled);
break;
case TaskExecutionMode.Parallel:
executor = new ParallelTaskExecutor();
fitness = new OptimizerFitness(StartDate, EndDate, fitScore, filterEnabled);
break;
case TaskExecutionMode.Azure:
executor = new ParallelTaskExecutor();
fitness = new AzureFitness(StartDate, EndDate, fitScore, filterEnabled);
break;
default:
throw new Exception("Executor initialization failed");
}
// Optimization mode
switch (Shared.Config.OptimizationMode)
{
case OptimizationMode.BruteForce:
{
// Create cartesian population
population = new PopulationCartesian(Shared.Config.GeneConfigArray);
termination = new GenerationNumberTermination(1);
break;
}
case OptimizationMode.Genetic:
{
// Create random population
population = new PopulationRandom(Shared.Config.GeneConfigArray, Shared.Config.PopulationInitialSize)
{
GenerationMaxSize = Shared.Config.GenerationMaxSize
};
// Logical terminaton
var localTerm = new LogicalOrTermination();
localTerm.AddTermination(new FruitlessGenerationsTermination(3));
if (Shared.Config.Generations.HasValue)
{
localTerm.AddTermination(new GenerationNumberTermination(Shared.Config.Generations.Value));
}
if (Shared.Config.StagnationGenerations.HasValue)
{
localTerm.AddTermination(new FitnessStagnationTermination(Shared.Config.StagnationGenerations.Value));
}
termination = localTerm;
break;
}
default:
throw new Exception("Optimization mode specific objects were not initialized");
}
// Create GA itself
GenAlgorithm = new GeneticAlgorithm(population, fitness, executor)
{
// Reference types
Selection = selection,
Termination = termination,
// Values types
CrossoverParentsNumber = Shared.Config.CrossoverParentsNumber,
CrossoverMixProbability = Shared.Config.CrossoverMixProbability,
MutationProbability = Shared.Config.MutationProbability
};
}
public GeneticOptimizer(GeneticOptimizerConfiguration configuration, Func <double[], double> objectiveFunction, Action <string> generationRanCallback = null)
{
//store configuration
_configuration = configuration;
_generationRanCallback = generationRanCallback;
//set min/max/precision of input variables
var minValues = new double[_configuration.Variables.Count];
var maxValues = new double[_configuration.Variables.Count];
var fractionDigits = new int[_configuration.Variables.Count];
for (int index = 0; index < _configuration.Variables.Count; index++)
{
minValues[index] = _configuration.Variables[index].MinimumValue;
maxValues[index] = _configuration.Variables[index].MaximumValue;
fractionDigits[index] = _configuration.Variables[index].NumberDigitsPrecision;
}
//total bits
var totalBits = new int[] { 64 };
//chromosome
var chromosome = new FloatingPointChromosome(minValues, maxValues, totalBits, fractionDigits);
//population
var population = new Population(MinimumNumberPopulation, MaximumNumberPopulation, chromosome);
//set fitness function
var fitnessFunction = new FuncFitness(c =>
{
var fc = c as FloatingPointChromosome;
var inputs = fc.ToFloatingPoints();
var result = objectiveFunction(inputs);
//add to results
if (!Double.IsNaN(result))
{
var list = inputs.ToList();
list.Add(result);
_result.IterationArray.Add(string.Join(",", list));
}
return(result);
});
//other variables
var selection = new EliteSelection();
var crossover = new UniformCrossover(0.5f);
var mutation = new FlipBitMutation();
var termination = new FitnessThresholdTermination();
_algorithm = new GeneticAlgorithm(population, fitnessFunction, selection, crossover, mutation)
{
Termination = termination,
};
//task parallelism
var taskExecutor = new ParallelTaskExecutor();
taskExecutor.MinThreads = 1;
taskExecutor.MaxThreads = _configuration.NumberThreadsToUse;
_algorithm.TaskExecutor = taskExecutor;
//if (_configuration.NumberThreadsToUse > 1)
//{
// var taskExecutor = new ParallelTaskExecutor();
// taskExecutor.MinThreads = 1;
// taskExecutor.MaxThreads = _configuration.NumberThreadsToUse;
// _algorithm.TaskExecutor = taskExecutor;
//}
//register generation ran callback
_algorithm.GenerationRan += AlgorithmOnGenerationRan;
}
