public GeneticLearner(int populationSize, int selectionSize, IModelInitializer modelInitializer, IModelExecuter modelExecuter, IPopulationBreeder populationBreeder, ThreadSafeRandom random)
{
this.populationSize = populationSize;
this.selectionSize = selectionSize;
this.modelInitializer = modelInitializer;
this.modelExecuter = modelExecuter;
this.populationBreeder = populationBreeder;
this.random = random;
populationCostLocks = new object[populationSize].Select(o => new object()).ToArray();
population = new CostModel <FullyConnectedNeuralNetworkModel> [populationSize];
}
public void TestInitialize()
{
random = new ThreadSafeRandom();
modelInitializer = new FullyConnectedNeuralNetworkInitializer(random);
sigmoidActivationFunction = new SigmoidActivationFunction();
modelExecuter = new FullyConnectedNeuralNetworkExecuter(sigmoidActivationFunction);
modelBreeder = new GeneticModelBreeder(modelInitializer, random);
populationBreeder = new PolygamousPopulationBreeder(modelBreeder, random);
sut = new GeneticLearner(populationSize, selectionSize, modelInitializer, modelExecuter, populationBreeder, random);
}
public Experiment(
string title,
int iterations,
int batchSize,
ActivationFunction activationFunction,
int[] activationCountsPerLayer,
IModelExecuter executer,
ILearner learner,
IModelInitializer modelInitializer,
IDataSource dataSource)
{
Title = title;
Iterations = iterations;
this.batchSize = batchSize;
this.activationFunction = activationFunction;
this.activationCountsPerLayer = activationCountsPerLayer;
this.executer = executer;
this.learner = learner;
this.modelInitializer = modelInitializer;
this.dataSource = dataSource;
random = new Random();
}
public BackPropagationLearner(IModelExecuter executer, [Named(nameof(SigmoidActivationFunction))] IActivationFunction sigmoidActivationFunction)
{
this.executer = executer;
this.sigmoidActivationFunction = sigmoidActivationFunction;
}
