static void Main(string[] args)
{
int[] array = { 2, 8, 4 };
string str = File.ReadAllText("TD1.txt");
float[][] vs1 = JsonConvert.DeserializeObject <float[][]>(str);
NNModel model = new NNModel(array);
BackPropagation backPropagation = new BackPropagation();
backPropagation.Learn(model, vs1);
for (int i = 0; i < vs1.Length; i += 2)
{
var inp = vs1[i];
Array.ForEach(model.Compute(inp).Neurons, x =>
{
Console.WriteLine(x.Value + " ");
});
Console.WriteLine();
}
Console.ReadKey();
}
public void Process()
{
Console.WriteLine("Please wait, reading MNIST training data.");
var dir = AppDomain.CurrentDomain.BaseDirectory;
var trainingReader = LearnDigitsBackprop.LoadMNIST(dir, true, MNIST_DEPTH);
var validationReader = LearnDigitsBackprop.LoadMNIST(dir, false, MNIST_DEPTH);
Console.WriteLine("Training set size: " + trainingReader.NumImages);
Console.WriteLine("Validation set size: " + validationReader.NumImages);
var inputCount = trainingReader.Data[0].Input.Length;
var outputCount = trainingReader.Data[0].Ideal.Length;
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(null, true, inputCount));
network.AddLayer(new BasicLayer(new ActivationReLU(), true, 100));
network.AddLayer(new DropoutLayer(new ActivationReLU(), true, 50, 0.5));
network.AddLayer(new BasicLayer(new ActivationReLU(), true, 25));
network.AddLayer(new BasicLayer(new ActivationSoftMax(), false, outputCount));
network.FinalizeStructure();
network.Reset();
// train the neural network
Console.WriteLine("Training neural network.");
var train = new BackPropagation(network, trainingReader.Data, 1e-4, 0.9);
train.L1 = 0;
train.L2 = 1e-11;
PerformIterationsClassifyEarlyStop(train, network, validationReader.Data, 5);
}
public ConfigurationFile()
{
// Retrieve instances
annComp = NinjectBinding.GetKernel.Get <AnnBuild>();
backPropComp = NinjectBinding.GetKernel.Get <BackPropagation>();
rPropComp = NinjectBinding.GetKernel.Get <ResilientPropagation>();
}
public void TrainNetwork()
{
List <int> list = new List <int>();
foreach (int layer in layers)
{
list.Add(layer);
}
Stream stream = assets.Open("Classroom_Occupation_Data.csv", Access.Random);
StreamReader sr = new StreamReader(stream);
string content = sr.ReadToEnd();
StreamWriter destinationStream = new StreamWriter(basePath + "/Classroom_Occupation_Data.csv", false);
destinationStream.Write(content);
destinationStream.Flush();
network = new MultiLayerPerceptron(TransferFunctionType.Sigmoid, list.ToArray());
trainingSet = new TrainingSet(inputSize, outputSize);
trainingSet = TrainingSet.CreateFromFile(basePath + "/Classroom_Occupation_Data.csv", inputSize, outputSize, ",");
BackPropagation learningRule = new BackPropagation();
network.LearningRule = learningRule;
network.Learn(trainingSet);
network.Save(basePath + "/TrainTest.nnet");
destinationStream.Close();
destinationStream.Dispose();
}
public BackPropagationPerformanceComparisonContainer()
{
var neuralNetworkUnderTest =
NeuralNetwork
.For(NeuralNetworkContext.MaximumPrecision)
.WithInputLayer(neuronCount: 5, activationType: ActivationType.Sigmoid)
.WithHiddenLayer(neuronCount: 100, activationType: ActivationType.Sigmoid)
.WithHiddenLayer(neuronCount: 70, activationType: ActivationType.TanH)
.WithHiddenLayer(neuronCount: 40, activationType: ActivationType.TanH)
.WithHiddenLayer(neuronCount: 100, activationType: ActivationType.Sigmoid)
.WithOutputLayer(neuronCount: 2, activationType: ActivationType.Sigmoid)
.Build();
trainingData = new[] {
TrainingDataSet.For(new [] { 0.78, 0.99, 0.67, 0.72, 0.22 }, new [] { 0.12, 0.14 })
};
multiThreadedController = TrainingController.For(
BackPropagation
.WithConfiguration(neuralNetworkUnderTest, ParallelOptionsExtensions.UnrestrictedMultiThreadedOptions));
singleThreadedController = TrainingController.For(
BackPropagation
.WithConfiguration(neuralNetworkUnderTest, ParallelOptionsExtensions.SingleThreadedOptions));
}
static void Testing()
{
int[] array = { 10, 20, 2 };
string str = File.ReadAllText("TD1.txt");
double[][] vs1 = JsonConvert.DeserializeObject <double[][]>(str);
NNModel model = new NNModel(array);
Console.Beep();
var watch = System.Diagnostics.Stopwatch.StartNew();
BackPropagation backPropagation = new BackPropagation();
backPropagation.Learn(model, vs1);
watch.Stop();
Console.WriteLine(watch.ElapsedMilliseconds);
string str1 = File.ReadAllText("TD2.txt");
double[][] vs2 = JsonConvert.DeserializeObject <double[][]>(str1);
for (int i = 0; i < vs2.Length; i++)
{
model.Compute(NNModel.ReculcArray(vs2[i])).Neurons.ForEach(x =>
{
Console.WriteLine(x.Value + " ");
});
Console.WriteLine();
}
Console.WriteLine("_______");
model.Save();
Console.Beep();
Console.ReadKey();
}
static void NewNN(string[] args)
{
int[] array = new int[args.Length - 1];
for (int i = 1; i < args.Length; i++)
{
if (Int32.TryParse(args[i], out int res))
{
array[i] = res;
}
else
{
return;
}
}
string str = File.ReadAllText(args[0]);
double[][] vs1 = JsonConvert.DeserializeObject <double[][]>(str);
NNModel model = new NNModel(array);
BackPropagation backPropagation = new BackPropagation();
backPropagation.Learn(model, vs1);
for (int i = 0; i < vs1.Length; i += 2)
{
var inp = vs1[i];
model.Compute(inp).Neurons.ForEach(x =>
{
Console.WriteLine(x.Value + " ");
});
Console.WriteLine();
}
Console.ReadKey();
}
public static void Main(string[] args)
{
//create training set from Data.DIGITS
DataSet dataSet = generateTraining();
int inputCount = Data.CHAR_HEIGHT * Data.CHAR_WIDTH;
int outputCount = Data.DIGITS.Length;
int hiddenNeurons = 19;
//create neural network
MultiLayerPerceptron neuralNet = new MultiLayerPerceptron(inputCount, hiddenNeurons, outputCount);
//get backpropagation learning rule from network
BackPropagation learningRule = neuralNet.LearningRule;
learningRule.LearningRate = 0.5;
learningRule.MaxError = 0.001;
learningRule.MaxIterations = 5000;
//add learning listener in order to print out training info
learningRule.addListener(new LearningEventListenerAnonymousInnerClassHelper());
//train neural network
neuralNet.learn(dataSet);
//train the network with training set
testNeuralNetwork(neuralNet, dataSet);
}
public static bool Test4()
{
var xorinput = new[] { new[] { 0.0, 0.0 }, new[] { 1.0, 0.0 }, new[] { 0.0, 1.0 }, new[] { 1.0, 1.0 } };
var xoroutput = new[] { new[] { 0.0 }, new[] { 1.0 }, new[] { 1.0 }, new[] { 0.0 } };
var mlp = new MLP(2, 4, 1);
var stsp = new StandardTrainingSetProvider(xorinput, xoroutput);
stsp.Split();
var gdbp = new BackPropagation(mlp, stsp);
((GD)gdbp.Solver).Criterion = LearningCriterion.CrossEntropy;
//((GD)gdbp.Solver).LearningRate = .01;
((GD)gdbp.Solver).AdaGrad = true;
var minerr = double.MaxValue;
gdbp.ReportReady += optimizer =>
{
Console.WriteLine("Epoch = {0}, Error = {1}", optimizer.CurrentEpoch,
optimizer.TrainingSetProvider.TrainError);
minerr = Math.Min(minerr, optimizer.Solver.Error);
if (optimizer.Done)
{
Console.ReadLine();
}
};
gdbp.RunAsync().Wait();
return(true);
}
/// <summary>
/// The entry point for this example. If you would like to make this example
/// stand alone, then add to its own project and rename to Main.
/// </summary>
/// <param name="args">Not used.</param>
public static void ExampleMain(string[] args)
{
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(null, true, 2));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 5));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, 1));
network.FinalizeStructure();
network.Reset();
var trainingData = BasicData.ConvertArrays(XOR_INPUT, XOR_IDEAL);
// train the neural network
var train = new BackPropagation(network, trainingData, 0.7, 0.9);
var epoch = 1;
do
{
train.Iteration();
Console.WriteLine("Epoch #" + epoch + " Error:" + train.LastError);
epoch++;
} while (train.LastError > 0.01);
// test the neural network
Console.WriteLine("Neural Network Results:");
for (var i = 0; i < XOR_INPUT.Length; i++)
{
var output = network.ComputeRegression(XOR_INPUT[i]);
Console.WriteLine(string.Join(",", XOR_INPUT[i])
+ ", actual=" + string.Join(",", output)
+ ",ideal=" + string.Join(",", XOR_IDEAL[i]));
}
}
public virtual void handleLearningEvent(LearningEvent @event)
{
BackPropagation bp = (BackPropagation)@event.Source;
Console.WriteLine("Current iteration: " + bp.CurrentIteration);
Console.WriteLine("Error: " + bp.TotalNetworkError);
}
static void TestANNMiner(Dataset dataset)
{
IClassificationMeasure measure = new AccuracyMeasure();
ILearningMethod learningMethod = new BackPropagation(0.1, 10, 0.9, false);
int hiddenUnitCount = dataset.Metadata.Attributes.Length * dataset.Metadata.Target.Length;
IActivationFunction activationFunction = new SigmoidActivationFunction();
ISolutionQualityEvaluator <ConnectionDC> evaluator = new NNClassificationQualityEvaluator(measure, learningMethod, hiddenUnitCount, activationFunction);
IHeuristicsCalculator <ConnectionDC> calculator = new DefaultHeuristicCalculator <ConnectionDC>();
ILocalSearch <ConnectionDC> localSearch = new DefaultRemovalLocalSearch <ConnectionDC>(evaluator);
IComponentInvalidator <ConnectionDC> invalidator = new NNConnectorInvalidator();
Problem <ConnectionDC> problem = new Problem <ConnectionDC>(invalidator, calculator, evaluator, localSearch);
NeuralNetwork network_before = null;
NeuralNetwork network_final = SingleTest.CreateNeuralNet_ANNMiner(problem, hiddenUnitCount, true, false, dataset, ref network_before);
double quilty_before = SingleTest.TestClassifier(network_before, dataset, measure);
double quilty_final = SingleTest.TestClassifier(network_final, dataset, measure);
Console.WriteLine("ANN -" + quilty_before);
Console.WriteLine("ANN -" + quilty_final);
}
void Start()
{
backpro = new BackPropagation(NumInput, NumHidden, NumOutput);
if (trained == false)
{
Train();
}
}
public void CreateBackPropagationTrainingAlgorithm(AnnBuild annComp, BackPropagation prop)
{
errorTarget = annComp.ErrorTarget;
network.InstantiateBackPropagationAlgorithm(
errorTarget, annComp.MaxEpochs, prop.LearningRate, prop.Beta);
network.BuildStatsListener();
}
static void Main(string[] args)
{
int[] layerSizes = new int[3] {
2, 2, 1
};
TransferFunction[] tFuncs = new TransferFunction[3] {
TransferFunction.None,
TransferFunction.Sigmoid,
TransferFunction.Linear
};
BackPropagation bpn = new BackPropagation(layerSizes, tFuncs);
double[][] input, output;
input = new double[4][]; output = new double[4][];
for (int i = 0; i < 4; i++)
{
input[i] = new double[2]; output[i] = new double[1];
}
input[0][0] = 0.0; input[0][1] = 0.0; output[0][0] = 0.0; // Case 1 F xor F = F
input[1][0] = 1.0; input[1][1] = 0.0; output[1][0] = 1.0; // Case 2 T xor F = T
input[2][0] = 0.0; input[2][1] = 1.0; output[2][0] = 1.0; // Case 3 F xor T = T
input[3][0] = 1.0; input[3][1] = 1.0; output[3][0] = 0.0; // Case 4 T xor T = F
// Train
double error = 0.0;
int max_count = 10000, count = 0;
do
{
// Prepare training epoch
count++;
error = 0.0;
// Train
for (int i = 0; i < 4; i++)
{
// TrainRate and Momentm picked arbitrarilly
error += bpn.Train(ref input[i], ref output[i], 0.15, 0.10);
}
// Show Progress
if (count % 100 == 0)
{
Console.WriteLine("Epoch {0} completed with error {1:0.0000}", count, error);
}
} while (error > 0.0001 && count <= max_count);
// Display results
double[] networkOutput = new double[1];
for (int i = 0; i < 4; i++)
{
bpn.Run(ref input[i], out networkOutput);
Console.WriteLine("Case {3}: {0:0.0} xor {1:0.0} = {2:0.0000}", input[i][0], input[i][1], networkOutput[0], i + 1);
}
// End Program
Console.WriteLine("Press Enter ...");
Console.ReadLine();
}
//creates an instance of the backpropagation with corresponding number of nodes and weights
void Start()
{
bpro = new BackPropagation(NumInput, NumHidden, NumOutput);
pass = false;
weights = new double[2451];
tint = GameObject.Find("playerInputs").GetComponent <throwinputs>();
Debug.Log("tssss");
}
public virtual void handleLearningEvent(LearningEvent @event)
{
BackPropagation bp = (BackPropagation)@event.Source;
if (@event.EventType != LearningEvent.Type.LEARNING_STOPPED)
{
Console.WriteLine(bp.CurrentIteration + ". iteration : " + bp.TotalNetworkError);
}
}
String GetCommandFromBackPropagation(int[,] sensorData, int sensorType)
{
int[,] processedSensorData = GetProcessedSensorData(sensorData);
var robotCommand =
BackPropagation.Driver("Command", _sensorTypeString[sensorType],
convertToOneDimensionalDouble(processedSensorData));
return(robotCommand);
}
private void BtnEntrenar_Click(object sender, EventArgs e)
{
if (TxtTipoRed.SelectedItem.ToString().Equals("Perceptrón Multicapa"))
{
percetronMulticapa = new PercetronMulticapa();
percetronMulticapa.NumeroEntradas = entradas;
percetronMulticapa.NumeroSalidas = salidas;
percetronMulticapa.NumeroPatrones = patrones;
percetronMulticapa.NumeroCapas = NumeroDeCapas;
percetronMulticapa.NumeroNeuronasCapas = ListaNumeroNeuronasCapas;
percetronMulticapa.FuncionesActivacionCapas = FuncionActivacionCapas;
percetronMulticapa.NumeroIteraciones = Convert.ToInt32(TxtNumeroIteraciones.Value.ToString());
percetronMulticapa.RataAprendizaje = Convert.ToDouble(TxtRataAprendizaje.Value.ToString());
percetronMulticapa.ErrorMaximoPermitido = Convert.ToDouble(TxtErrorMaximoPermitido.Value.ToString());
percetronMulticapa.FuncionActivacionCapaSalida = CbxFuncionActivacionCapaSalida.SelectedItem.ToString();
percetronMulticapa.Umbrales = arraydeumbrales;
percetronMulticapa.Pesos = arraydematrizdepesos;
}
else
{
backPropagation = new BackPropagation();
backPropagation.NumeroEntradas = entradas;
backPropagation.NumeroSalidas = salidas;
backPropagation.NumeroSalidas = patrones;
backPropagation.NumeroCapas = NumeroDeCapas;
backPropagation.NumeroNeuronasCapas = ListaNumeroNeuronasCapas;
backPropagation.FuncionesActivacionCapas = FuncionActivacionCapas;
backPropagation.NumeroIteraciones = Convert.ToInt32(TxtNumeroIteraciones.Value.ToString());
backPropagation.RataAprendizaje = Convert.ToDouble(TxtRataAprendizaje.Value.ToString());
backPropagation.ErrorMaximoPermitido = Convert.ToDouble(TxtErrorMaximoPermitido.Value.ToString());
backPropagation.FuncionActivacionCapaSalida = CbxFuncionActivacionCapaSalida.SelectedItem.ToString();
backPropagation.Umbrales = arraydeumbrales;
backPropagation.Pesos = arraydematrizdepesos;
MessageBox.Show("Numero entradas: " + backPropagation.NumeroEntradas.ToString() + "" +
"NumeroSalidas: " + backPropagation.NumeroSalidas.ToString() + "" +
"NumeroPatrones: " + backPropagation.NumeroPatrones.ToString() + "" +
"NumeroCapas: "******"" +
"NumeroNeuronasCapas: "******"" +
"FuncionActivacionCapas: "******"" +
"NumeroIteraciones: " + backPropagation.NumeroIteraciones.ToString() + "" +
"RataAprendizaje: " + backPropagation.RataAprendizaje.ToString() + "" +
"ErrorMaximoPermitido: " + backPropagation.ErrorMaximoPermitido.ToString() + "" +
"FuncionActivacionCapaSalida: " + backPropagation.FuncionActivacionCapaSalida.ToString() + "" +
"Umbrales: " + backPropagation.Umbrales[0].ToString() + "" +
"Pesos: " + backPropagation.Pesos[0, 0].ToString() + "", "algo");
}
}
public double Train(double[] input, double[] output, BackPropagation trainer)
{
// Run the given input throug the network in order to obtain output
m_Network.Run (input);
// Calculate network error between the generated output and the expected output
double error = CalculateError (output);
// Update the network to remedy the calculated error
Update (input, trainer.LearningRate, trainer.Momentum);
return error;
}
private void testWithParametr(AlgorithmConfig configuration, Vector[] learnVectors, int classes)
{
neuronsCount = null;
int layersCount = (int)hiddenLayers.Value + 1;
switch (layersCount)
{
case 2:
{
neuronsCount = new int[1];
neuronsCount[0] = (int)firstLayerNeurons.Value;
} break;
case 3:
{
neuronsCount = new int[2];
neuronsCount[0] = (int)firstLayerNeurons.Value;
neuronsCount[1] = (int)secondtLayerNeurons.Value;
} break;
case 4:
{
neuronsCount = new int[3];
neuronsCount[0] = (int)firstLayerNeurons.Value;
neuronsCount[1] = (int)secondtLayerNeurons.Value;
neuronsCount[2] = (int)thirdtLayerNeurons.Value;
} break;
}
beta = 0.7;
myNet = new NeuralNetwork(layersCount, neuronsCount, learnVectors[0].input.Length, classes, beta);
List <Vector> data = new List <Vector>(learnVectors.Length);
data.AddRange(learnVectors);
switch (algorithmBox.SelectedIndex)
{
case 0:
{
BackPropagation method = new BackPropagation(configuration);
method.RaiseMessageEvent += AddMessage;
method.Train(myNet, data);
break;
}
case 1:
{
QuickProp method = new QuickProp(configuration);
method.RaiseMessageEvent += AddMessage;
method.Train(myNet, data);
break;
}
}
}
public BackPropagationTester QueueTrainingEpoch(
Action <TrainingEpochTester <BackPropagation> > action,
ParallelOptions parallelOptions)
{
var backPropagationTrainer = BackPropagation.WithConfiguration(targetNeuralNetwork, parallelOptions, learningRate, momentum);
var epochTester = TrainingEpochTester <BackPropagation> .For(backPropagationTrainer);
action.Invoke(epochTester);
trainingStates.Add(epochTester);
return(this);
}
public virtual void handleLearningEvent(LearningEvent @event)
{
BackPropagation bp = (BackPropagation)@event.Source;
LOG.info("Epoch no#: [{}]. Error [{}]", bp.CurrentIteration, bp.TotalNetworkError);
LOG.info("Epoch execution time: {} sec", (DateTimeHelperClass.CurrentUnixTimeMillis() - start) / 1000.0);
// neuralNetwork.save(bp.getCurrentIteration() + "_MNIST_CNN-MIC.nnet");
start = DateTimeHelperClass.CurrentUnixTimeMillis();
// if (bp.getCurrentIteration() % 5 == 0)
// Evaluation.runFullEvaluation(neuralNetwork, testSet);
}
public BackPropagation GetLearningMethod(object owner, IContextLookup globalVars, BasicNetwork theNetwork,
IList <BasicData> data)
{
double theLearningRate = LearningRate.GetValue(owner, globalVars);
double theMomentum = Momentum.GetValue(owner, globalVars);
var theL1 = L1.GetValue(owner, globalVars);
var theL2 = L2.GetValue(owner, globalVars);
var toReturn = new BackPropagation(theNetwork, data, theLearningRate, theMomentum);
toReturn.L1 = theL1;
toReturn.L2 = theL2;
return(toReturn);
}
private void TrainNeuralNetworks(int algorithmType, int sensorType)
{
switch (algorithmType)
{
case (int)AlgorithmType.BackPropagation:
BackPropagation.Driver("Train", _sensorTypeString[sensorType], null);
break;
case (int)AlgorithmType.FeedForward:
forwardManager.GetDirectionFromFeedForward(_sensorTypeString[sensorType], new float[] { }, 0);
break;
}
}
private void UczenieSieci_Click(object sender, RoutedEventArgs e)
{
myVector x = bow.GetVectorsList()[0];
BackPropagation.UczenieSieci(200, test.GetTrainingtVectors(), NeuralNetwork, classes);
Console.Beep();
foreach (myVector testvector in test.GetTestVectors())
{
int id = NeuralConstruction.SampleInput(testvector, NeuralNetwork);
var output = NeuralNetwork.getNetwork().Where(o => o.type == 2).ToList();
}
UczenieSieci.IsEnabled = false;
}
public virtual void handleLearningEvent(LearningEvent @event)
{
BackPropagation bp = (BackPropagation)@event.Source;
if (@event.EventType.Equals(LearningEvent.Type.LEARNING_STOPPED))
{
double error = bp.TotalNetworkError;
Console.WriteLine("Training completed in " + bp.CurrentIteration + " iterations, ");
Console.WriteLine("With total error: " + formatDecimalNumber(error));
}
else
{
Console.WriteLine("Iteration: " + bp.CurrentIteration + " | Network error: " + bp.TotalNetworkError);
}
}
public async Task TrainConfiguredNetworkForEpochs(
int epochs,
NeuralNetworkTrainingConfiguration trainingConfig)
{
var trainingData = dataProvider.TrainingData
.Select(transaction => transaction.ToTrainingData())
.ToList();
await TrainingController.For(BackPropagation.WithConfiguration(
networkAccessor.TargetNetwork,
ParallelOptionsExtensions.MultiThreadedOptions(trainingConfig.ThreadCount),
trainingConfig.LearningRate,
trainingConfig.Momentum))
.TrainForEpochsOrErrorThresholdMet(trainingData, epochs, trainingConfig.MinimumErrorThreshold);
}
public static NeuralNetwork CreateNeuralNet_BP(Dataset trainingset, int hiddenUnitCount, double positiveClassValue, double learningRate, int epochCount, IActivationFunction activationFunction)
{
counter = 0;
//MSError measure = new MSError();
//AccuracyMeasure measure = new AccuracyMeasure();
Connection[] connections = NeuralNetwork.Create3LayerConnectedConnections(trainingset.Metadata, hiddenUnitCount);
NeuralNetwork network = new NeuralNetwork(trainingset.Metadata, hiddenUnitCount, activationFunction, connections);
BackPropagation BP = new BackPropagation(learningRate, epochCount, positiveClassValue, true);
BP.OnPostEpoch += new EventHandler(OnPostEpoch);
BP.TrainNetwork(network, trainingset);
return(network);
}
public async void CanSuccessfullySolveXorProblemTrainingForErrorThreshold()
{
var neuralNetwork = NeuralNetwork.For(NeuralNetworkContext.MaximumPrecision)
.WithInputLayer(neuronCount: 2, activationType: ActivationType.Sigmoid)
.WithHiddenLayer(neuronCount: 2, activationType: ActivationType.TanH)
.WithOutputLayer(neuronCount: 1, activationType: ActivationType.Sigmoid)
.Build();
await TrainingController
.For(BackPropagation.WithConfiguration(
neuralNetwork,
ParallelOptionsExtensions.UnrestrictedMultiThreadedOptions,
learningRate: 0.4,
momentum: 0.9))
.TrainForErrorThreshold(XorTrainingData(), minimumErrorThreshold: 0.01);
AssertPredictionsForTrainedNeuralNetwork(neuralNetwork);
}
public async void CanSuccessfullySolveIrisProblemTrainingForEpochs()
{
var neuralNetwork = NeuralNetwork.For(NeuralNetworkContext.MaximumPrecision)
.WithInputLayer(neuronCount: 4, activationType: ActivationType.Sigmoid)
.WithHiddenLayer(neuronCount: 8, activationType: ActivationType.Sigmoid)
.WithHiddenLayer(neuronCount: 5, activationType: ActivationType.Sigmoid)
.WithOutputLayer(neuronCount: 3, activationType: ActivationType.TanH)
.Build();
await TrainingController
.For(BackPropagation.WithConfiguration(
neuralNetwork,
ParallelOptionsExtensions.UnrestrictedMultiThreadedOptions,
learningRate: 1.15,
momentum: 0.4))
.TrainForEpochs(IrisDataSet.TrainingData, maximumEpochs: 1000);
AssertPredictionsForTrainedNeuralNetwork(neuralNetwork);
}
IEnumerator Start()
{
m_Running = true;
Debug.Log ("Starting test");
yield return null;
Debug.Log ("Setup");
m_Network = new Network (InputCount, LayerCounts);
if (m_Randomize)
{
m_Network.RandomizeWeights ();
}
BackPropagation teacher = new BackPropagation (m_Network)
{
LearningRate = this.LearningRate,
Momentum = this.Momentum
};
if (m_LearningCurve != null)
{
m_LearningCurve.Curve.keys = new Keyframe[] {};
}
if (m_Step)
{
Debug.Break ();
}
yield return null;
Debug.Log ("Start training");
while (m_Running && enabled && Application.isPlaying)
{
double error = teacher.TrainSet (Input, Output);
LogResult ("Training run error: " + error);
if (m_LearningCurve != null)
{
m_LearningCurve.Curve.AddKey (Time.time, (float)error);
}
if (error < m_TargetError)
{
Debug.Log (string.Format ("Achieved error {0}, beating target {1}", error, m_TargetError));
yield break;
}
if (m_Step)
{
Debug.Break ();
}
yield return null;
}
}
