public void testDataSetPopulation()
{
DataSet irisDataSet = DataSetFactory.getIrisDataSet();
INumerizer numerizer = new IrisDataSetNumerizer();
NeuralNetworkDataSet innds = new IrisNeuralNetworkDataSet();
innds.CreateExamplesFromDataSet(irisDataSet, numerizer);
NeuralNetworkConfig config = new NeuralNetworkConfig();
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_INPUTS, 4);
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_OUTPUTS, 3);
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_HIDDEN_NEURONS, 6);
config.SetConfig(FeedForwardNeuralNetwork.LOWER_LIMIT_WEIGHTS, -2.0);
config.SetConfig(FeedForwardNeuralNetwork.UPPER_LIMIT_WEIGHTS, 2.0);
FeedForwardNeuralNetwork ffnn = new FeedForwardNeuralNetwork(config);
ffnn.SetTrainingScheme(new BackPropagationLearning(0.1, 0.9));
ffnn.TrainOn(innds, 10);
innds.RefreshDataset();
ffnn.TestOnDataSet(innds);
}
internal static void backPropogationDemo()
{
try
{
DataSet irisDataSet = DataSetFactory.getIrisDataSet();
INumerizer numerizer = new IrisDataSetNumerizer();
NeuralNetworkDataSet innds = new IrisNeuralNetworkDataSet();
innds.CreateExamplesFromDataSet(irisDataSet, numerizer);
NeuralNetworkConfig config = new NeuralNetworkConfig();
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_INPUTS, 4);
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_OUTPUTS, 3);
config.SetConfig(FeedForwardNeuralNetwork.NUMBER_OF_HIDDEN_NEURONS,
6);
config.SetConfig(FeedForwardNeuralNetwork.LOWER_LIMIT_WEIGHTS, -2.0);
config.SetConfig(FeedForwardNeuralNetwork.UPPER_LIMIT_WEIGHTS, 2.0);
FeedForwardNeuralNetwork ffnn = new FeedForwardNeuralNetwork(config);
ffnn.SetTrainingScheme(new BackPropagationLearning(0.1, 0.9));
ffnn.TrainOn(innds, 1000);
innds.RefreshDataset();
int[] result = ffnn.TestOnDataSet(innds);
System.Console.WriteLine(result[0] + " right, " + result[1] + " wrong");
}
catch (Exception e)
{
throw e;
}
}
public static NeuralNetwork GetBrain()
{
if (_brain == null)
{
NeuralNetworkConfig nnConfig = new NeuralNetworkConfig();
nnConfig.AddLayer(LayerType.Input, 5); //Vision inouts
nnConfig.AddLayer(LayerType.Intermediate, 12, ActivationType.LeakyRelu);
nnConfig.AddLayer(LayerType.Output, 5, ActivationType.LeakyRelu); //Movement decisions.
_brain = new NeuralNetwork(nnConfig, 0.02f);
for (int i = 0; i < 10000; i++)
{
//Left side detection, go right.
_brain.BackPropagate(TrainingScenerio(1, 0, 0, 0, 0), TrainingDecision(1, 1, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(0, 1, 0, 0, 0), TrainingDecision(1, 1, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(1, 1, 0, 0, 0), TrainingDecision(1, 1, 1, 1, 0));
//Right side detection, go left.
_brain.BackPropagate(TrainingScenerio(0, 0, 0, 0, 1), TrainingDecision(1, 0, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(0, 0, 0, 1, 0), TrainingDecision(1, 0, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(0, 0, 0, 1, 1), TrainingDecision(1, 0, 1, 1, 0));
//Front side detection, so left or right.
_brain.BackPropagate(TrainingScenerio(0, 0, 1, 0, 0), TrainingDecision(1, 0, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(0, 1, 1, 1, 0), TrainingDecision(1, 1, 1, 1, 0));
_brain.BackPropagate(TrainingScenerio(1, 1, 1, 1, 1), TrainingDecision(1, 1, 1, 1, 0));
//No objects dection, speed up and cruise.
_brain.BackPropagate(TrainingScenerio(0, 0, 0, 0, 0), TrainingDecision(0.4f, 0.4f, 0.4f, 0.9f, 0.9f));
}
}
return _brain.Clone();
}
private static void Start()
{
NeuralNetworkConfig nnConfig = new NeuralNetworkConfig();
nnConfig.AddLayer(LayerType.Input, 3);
nnConfig.AddLayer(LayerType.Intermediate, 5, ActivationType.LeakyRelu);
nnConfig.AddLayer(LayerType.Output, 1, ActivationType.LeakyRelu);
nn = new NeuralNetwork(nnConfig, 0.01f);
nn.Load("C:\\network.txt");
//Train the network
for (int i = 0; i < 20000; i++)
{
nn.BackPropagate(new float[] { 0, 0, 0 }, new float[] { 0 });
nn.BackPropagate(new float[] { 1, 0, 0 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 1, 0 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 0, 1 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 0, 1 }, new float[] { 0 });
nn.BackPropagate(new float[] { 0, 0, 1 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 0, 1 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 0, 1 }, new float[] { 0 });
nn.BackPropagate(new float[] { 1, 1, 0 }, new float[] { 1 });
nn.BackPropagate(new float[] { 0, 1, 1 }, new float[] { 1 });
nn.BackPropagate(new float[] { 1, 0, 1 }, new float[] { 1 });
nn.BackPropagate(new float[] { 1, 1, 1 }, new float[] { 1 });
}
Console.WriteLine($"Cost: {nn.cost:0.########}");
nn.Save("C:\\network.txt");
VerboseFeedForward(nn, new float[] { 0, 0, 0 });
VerboseFeedForward(nn, new float[] { 1, 0, 0 });
VerboseFeedForward(nn, new float[] { 0, 1, 0 });
VerboseFeedForward(nn, new float[] { 0, 0, 1 });
VerboseFeedForward(nn, new float[] { 1, 1, 0 });
VerboseFeedForward(nn, new float[] { 0, 1, 1 });
VerboseFeedForward(nn, new float[] { 1, 0, 1 });
VerboseFeedForward(nn, new float[] { 1, 1, 1 });
//We want the gate to simulate 3 input or gate (A or B or C)
// 0 0 0 => 0
// 1 0 0 => 1
// 0 1 0 => 1
// 0 0 1 => 1
// 1 1 0 => 1
// 0 1 1 => 1
// 1 0 1 => 1
// 1 1 1 => 1
}
public NeuralNetworkEngineTests()
{
_config400x25x10 = new NeuralNetworkConfig
{
InputUnits = 400,
HiddenLayers = new int[] { 25 },
OutputUnits = 10
};
_config1x2X2x1 = new NeuralNetworkConfig
{
InputUnits = 1,
HiddenLayers = new int[] { 2, 2 },
OutputUnits = 1
};
}
internal static void backPropogationDeepLearningDemo()
{
try
{
System.Console.WriteLine(Util.ntimes("*", 100));
System.Console.WriteLine(
"\n BackpropagationnDemo - Running BackProp {1} hidden layers on Iris data Set with {0} epochs of learning ",
epochs, numHiddenLayers);
System.Console.WriteLine(Util.ntimes("*", 100));
DataSet animalDataSet = DataSetFactory.getAnimalDataSet();
INumerizer numerizer = new AnimalDataSetNumerizer();
NeuralNetworkDataSet innds = new IrisNeuralNetworkDataSet();
innds.CreateExamplesFromDataSet(animalDataSet, numerizer);
NeuralNetworkConfig config = new NeuralNetworkConfig();
config.SetConfig(FeedForwardDeepNeuralNetwork.NUMBER_OF_INPUTS, 20);
config.SetConfig(FeedForwardDeepNeuralNetwork.NUMBER_OF_OUTPUTS, 3);
config.SetConfig(FeedForwardDeepNeuralNetwork.NUMBER_OF_HIDDEN_LAYERS, numHiddenLayers);
config.SetConfig(FeedForwardDeepNeuralNetwork.NUMBER_OF_HIDDEN_NEURONS_PER_LAYER, numNeuronsPerLayer);
config.SetConfig(FeedForwardDeepNeuralNetwork.LOWER_LIMIT_WEIGHTS, -2.0);
config.SetConfig(FeedForwardDeepNeuralNetwork.UPPER_LIMIT_WEIGHTS, 2.0);
FeedForwardDeepNeuralNetwork ffnn = new FeedForwardDeepNeuralNetwork(config, new SoftSignActivationFunction());
ffnn.SetTrainingScheme(new BackPropagationDeepLearning(0.1, 0.9));
ffnn.TrainOn(innds, epochs);
innds.RefreshDataset();
int[] result = ffnn.TestOnDataSet(innds);
System.Console.WriteLine(result[0] + " right, " + result[1] + " wrong");
}
catch (Exception e)
{
throw e;
}
}
