/// <summary>
/// Create, Train and Test a Garden Design Network
/// ToDo: Add Dependency Injection
/// </summary>
private static void CreateGardenDesignNetwork()
{
//Create the Network
INeuralNetworkFactory neuralNetworkFactory = new NeuralNetworkFactory();
var neuralNetwork = neuralNetworkFactory.Create(3, 5, 3);
// Train the Network
ITrainingDataRepo trainingDataRepo = new TrainingDataRepo();
var trainingData = trainingDataRepo.GetTrainingDataGardenDesign();
INeuralNetworkTrainingService trainingService = new NeuralNetworkTrainingService(new NeuralNetworkProcessingService());
double[][] trainingResults = trainingService.Train(neuralNetwork, trainingData, 4000, new Sigmoid(), 1);
OutputResultsToConsole(trainingResults);
Console.WriteLine("Training Complete ---------------");
// Test the Trained Network
INeuralNetworkProcessingService processingService = new NeuralNetworkProcessingService();
double[] testInputs1 = { 0, 0, 0 };
double[][] testResults1 = { testInputs1, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs1, new Sigmoid()) };
Console.WriteLine("Test1 Results (0, 0, 0 Low Maint(1)) ---------------");
OutputResultsToConsole(testResults1);
double[] testInputs2 = { 1, 1, 0 };
double[][] testResults2 = { testInputs2, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs2, new Sigmoid()) };
Console.WriteLine("Test2 Results (1, 1, 0 General(5)) ---------------");
OutputResultsToConsole(testResults2);
double[] testInputs3 = { 0, 0, 1 };
double[][] testResults3 = { testInputs3, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs3, new Sigmoid()) };
Console.WriteLine("Test3 Results (0, 0, 1 Child(4)) ---------------");
OutputResultsToConsole(testResults3);
}
public INeuralNetwork LoadNeuralNetwork(string filename)
{
var jsonGenes = File.ReadAllText(_directory + filename);
var networkGene = JsonConvert.DeserializeObject <NeuralNetworkGene>(jsonGenes);
var network = NeuralNetworkFactory.GetInstance().Create(networkGene);
return(network);
}
internal ITrainingSession ExtractSession(string serialized, double eval)
{
var networkGenes = JsonConvert.DeserializeObject <NeuralNetworkGene>(serialized);
var network = NeuralNetworkFactory.GetInstance().Create(networkGenes);
var session = new FakeTrainingSession(network, eval);
return(session);
}
public static IGeneticAlgorithmFactory GetInstance(IEvaluatableFactory evaluatableFactory)
{
var networkFactory = NeuralNetworkFactory.GetInstance();
var workingSetFactory = EvalWorkingSetFactory.GetInstance();
var random = new Random();
var breederFactory = BreederFactory.GetInstance(networkFactory, new RandomWeightInitializer(random));
var mutatorFactory = MutatorFactory.GetInstance(networkFactory, new RandomWeightInitializer(random));
return(new GeneticAlgorithmFactory(networkFactory, workingSetFactory, evaluatableFactory, breederFactory, mutatorFactory));
}
// Use this for initialization
void Start()
{
createPopulation();
l_og.invokeGenerateObstacle();
lastFitnessTop1Bird = 0;
lastFitnessTop2Bird = 0;
l_neuralNetworkFactory = new NeuralNetworkFactory();
}
public async Task <ITrainingSession> GetBestSessionAsync()
{
var result = await ParseCloud.CallFunctionAsync <ParseObject>("bestNetwork", new Dictionary <string, object> {
{ "networkVersion", _networkVersion }
});
var networkGenes = JsonConvert.DeserializeObject <NeuralNetworkGene>((string)result["jsonNetwork"]);
var network = NeuralNetworkFactory.GetInstance().Create(networkGenes);
var session = new FakeTrainingSession(network, (double)result["eval"]);
return(session);
}
static void Main(string[] args)
{
NeuralNetworkConfigurationSettings networkConfig = new NeuralNetworkConfigurationSettings
{
NumInputNeurons = 1,
NumOutputNeurons = 1,
NumHiddenLayers = 2,
NumHiddenNeurons = 3,
SummationFunction = new SimpleSummation(),
ActivationFunction = new TanhActivationFunction()
};
GenerationConfigurationSettings generationSettings = new GenerationConfigurationSettings
{
UseMultithreading = true,
GenerationPopulation = 500
};
EvolutionConfigurationSettings evolutionSettings = new EvolutionConfigurationSettings
{
NormalMutationRate = 0.05,
HighMutationRate = 0.5,
GenerationsPerEpoch = 10,
NumEpochs = 1000,
NumTopEvalsToReport = 10
};
MutationConfigurationSettings mutationSettings = new MutationConfigurationSettings
{
MutateAxonActivationFunction = true,
MutateNumberOfHiddenLayers = true,
MutateNumberOfHiddenNeuronsInLayer = true,
MutateSomaBiasFunction = true,
MutateSomaSummationFunction = true,
MutateSynapseWeights = true
};
var random = new RandomWeightInitializer(new Random());
INeuralNetworkFactory factory = NeuralNetworkFactory.GetInstance(SomaFactory.GetInstance(networkConfig.SummationFunction), AxonFactory.GetInstance(networkConfig.ActivationFunction), SynapseFactory.GetInstance(new RandomWeightInitializer(new Random()), AxonFactory.GetInstance(networkConfig.ActivationFunction)), SynapseFactory.GetInstance(new ConstantWeightInitializer(1.0), AxonFactory.GetInstance(new IdentityActivationFunction())), random);
IBreeder breeder = BreederFactory.GetInstance(factory, random).Create();
IMutator mutator = MutatorFactory.GetInstance(factory, random).Create(mutationSettings);
IEvalWorkingSet history = EvalWorkingSetFactory.GetInstance().Create(50);
IEvaluatableFactory evaluatableFactory = new GameEvaluationFactory();
IStorageProxy proxy = new NodeJSProxy(1, "http://localhost:3000", "123456789");
IEpochAction action = new BestPerformerUpdater(proxy);
var GAFactory = GeneticAlgorithmFactory.GetInstance(evaluatableFactory);
IGeneticAlgorithm evolver = GAFactory.Create(networkConfig, generationSettings, evolutionSettings, factory, breeder, mutator, history, evaluatableFactory, action);
evolver.RunSimulation();
}
/// <summary>
/// Create, Train and Test an XOR Network
/// ToDo: Add Dependency Injection
/// </summary>
private static void CreateXORNetwork()
{
//Create the Network
// Initiate Weights
double[] hiddenLayerWeights = new double[] { 0.38286575972236, 0.441739972420847, 0.00519391335788831, 0.152334728814817 };
double[] hiddenLayerBias = new double[] { 0.79060369394282, 0.25095684418965 };
double[] outputLayerWeights = new double[] { 0.100294556049767, 0.24866028653861 };
double[] outputLayerBias = new double[] { 0.97103805186741 };
INeuralNetworkFactory neuralNetworkFactory = new NeuralNetworkFactory();
var neuralNetwork = neuralNetworkFactory.Create(2, 1, 2, hiddenLayerWeights, outputLayerWeights, hiddenLayerBias, outputLayerBias);
// Train the Network
ITrainingDataRepo trainingDataRepo = new TrainingDataRepo();
var trainingData = trainingDataRepo.GetTrainingDataXOR();
INeuralNetworkTrainingService trainingService = new NeuralNetworkTrainingService(new NeuralNetworkProcessingService());
double[][] trainingResults = trainingService.Train(neuralNetwork, trainingData, 2000, new Sigmoid(), 1);
OutputResultsToConsole(trainingResults);
Console.WriteLine("Training Complete ---------------");
// Test the Trained Network
INeuralNetworkProcessingService processingService = new NeuralNetworkProcessingService();
double[] testInputs1 = { 0, 0 };
double[][] testResults1 = { testInputs1, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs1, new Sigmoid()) };
Console.WriteLine("Test1 Results (0 XOR 0) ---------------");
OutputResultsToConsole(testResults1);
double[] testInputs2 = { 1, 0 };
double[][] testResults2 = { testInputs2, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs2, new Sigmoid()) };
Console.WriteLine("Test2 Results (1 XOR 0) ---------------");
OutputResultsToConsole(testResults2);
double[] testInputs3 = { 0, 1 };
double[][] testResults3 = { testInputs3, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs3, new Sigmoid()) };
Console.WriteLine("Test3 Results (0 XOR 1) ---------------");
OutputResultsToConsole(testResults3);
double[] testInputs4 = { 1, 1 };
double[][] testResults4 = { testInputs4, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs4, new Sigmoid()) };
Console.WriteLine("Test4 Results (1 XOR 1) ---------------");
OutputResultsToConsole(testResults4);
}
public INeuralNetworkFactory BuildBackpropagationNetworkFactory(IWeightInitializer weightInitializer,
ISomaFactory somaFactory,
IActivationFunctionDerivative activationFunctionDerivative,
IActivationFunction inputActivationFunction,
INeuronFactory neuronFactory)
{
var axonFactory = BackpropagationAxonFactory.GetInstance(activationFunctionDerivative);
var hiddenSynapseFactory = DecoratedSynapseFactory.GetInstance(weightInitializer,
AxonFactory.GetInstance(activationFunctionDerivative));
var inputSynapseFactory = DecoratedSynapseFactory.GetInstance(new ConstantWeightInitializer(1.0),
AxonFactory.GetInstance(inputActivationFunction));
var decoratedNeuronFactory = BackpropagationNeuronFactory.GetInstance(neuronFactory);
INeuralNetworkFactory factory = NeuralNetworkFactory.GetInstance(somaFactory, axonFactory,
hiddenSynapseFactory, inputSynapseFactory, weightInitializer, decoratedNeuronFactory);
var backPropNetworkFactory = new BackpropagationNetworkFactoryDecorator(factory);
return(backPropNetworkFactory);
}
public ITrainingSession GetBestSession()
{
using (var client = new HttpClient())
{
// Replace this with the API key for the web service
client.DefaultRequestHeaders.Authorization = new AuthenticationHeaderValue("Bearer", _apiKey);
var networkEndpoint = _baseUrl + "/network/" + _networkVersion;
var response = client.GetAsync(networkEndpoint).Result;
var responseObj = response.Content.ReadAsAsync <NodeJsMessage>().Result;
if (responseObj == null)
{
return(null);
}
var networkGenes = JsonConvert.DeserializeObject <NeuralNetworkGene>(responseObj.networkGenes);
var network = NeuralNetworkFactory.GetInstance().Create(networkGenes);
var session = new FakeTrainingSession(network, responseObj.eval);
return(session);
}
}
/// <summary>
/// Create, Train and Test an AND Network
/// ToDo: Add Dependency Injection
/// </summary>
private static void CreateANDNetwork()
{
//Create the Network (go with random weights and bias)
INeuralNetworkFactory neuralNetworkFactory = new NeuralNetworkFactory();
var neuralNetwork = neuralNetworkFactory.Create(2, 1, 2);
// Train the Network
ITrainingDataRepo trainingDataRepo = new TrainingDataRepo();
var trainingData = trainingDataRepo.GetTrainingDataAND();
INeuralNetworkTrainingService trainingService = new NeuralNetworkTrainingService(new NeuralNetworkProcessingService());
double[][] trainingResults = trainingService.Train(neuralNetwork, trainingData, 2000, new Sigmoid(), 1);
OutputResultsToConsole(trainingResults);
Console.WriteLine("Training Complete ---------------");
// Test the Trained Network
INeuralNetworkProcessingService processingService = new NeuralNetworkProcessingService();
double[] testInputs1 = { 0, 0 };
double[][] testResults1 = { testInputs1, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs1, new Sigmoid()) };
Console.WriteLine("Test1 Results (0 AND 0) ---------------");
OutputResultsToConsole(testResults1);
double[] testInputs2 = { 1, 0 };
double[][] testResults2 = { testInputs2, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs2, new Sigmoid()) };
Console.WriteLine("Test2 Results (1 AND 0) ---------------");
OutputResultsToConsole(testResults2);
double[] testInputs3 = { 0, 1 };
double[][] testResults3 = { testInputs3, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs3, new Sigmoid()) };
Console.WriteLine("Test3 Results (0 AND 1) ---------------");
OutputResultsToConsole(testResults3);
double[] testInputs4 = { 1, 1 };
double[][] testResults4 = { testInputs4, processingService.ProcessFeedForwardNeuralNetwork(neuralNetwork, testInputs4, new Sigmoid()) };
Console.WriteLine("Test4 Results (1 AND 1) ---------------");
OutputResultsToConsole(testResults4);
}
