public void DFFNeuralNetwork_InitializesLayers()
{
// Arrange
var network = new DFFNeuralNetwork();
// Assert
Assert.IsNotNull(network.Layers);
}
public void ApplyInputs_WithValidNetwork_PropogatesNetworkInputs()
{
// Arrange
var network = new DFFNeuralNetwork(1, 1, 2, 2);
var inputs = new List <INetworkInput>()
{
new NetworkInput()
{
ActivationLevel = .75
}
};
network.RandomizeNetwork();
IInputNeuron input1 = network.Layers.OfType <IInputLayer>().First().Neurons.First() as IInputNeuron;
IOutgoingConnection input1Hidden1Conn = input1.Connections.OfType <IOutgoingConnection>().First();
IOutgoingConnection input1Hidden2Conn = input1.Connections.OfType <IOutgoingConnection>().ToList()[1];
IHiddenNeuron hidden1 = input1Hidden1Conn.ToNeuron as IHiddenNeuron;
IOutgoingConnection hidden1Output1Conn = hidden1.Connections.OfType <IOutgoingConnection>().First();
IOutgoingConnection hidden1Output2Conn = hidden1.Connections.OfType <IOutgoingConnection>().ToList()[1];
IHiddenNeuron hidden2 = input1Hidden2Conn.ToNeuron as IHiddenNeuron;
IOutgoingConnection hidden2Output1Conn = hidden2.Connections.OfType <IOutgoingConnection>().First();
IOutgoingConnection hidden2Output2Conn = hidden2.Connections.OfType <IOutgoingConnection>().ToList()[1];
IOutputNeuron output1 = hidden1Output1Conn.ToNeuron as IOutputNeuron;
IOutputNeuron output2 = hidden1Output2Conn.ToNeuron as IOutputNeuron;
// Act
network.ApplyInputs(inputs);
var outputLayer = network.Layers.OfType <IOutputLayer>().First();
IOutputNeuron output1After = outputLayer.Neurons.First() as IOutputNeuron;
IOutputNeuron output2After = outputLayer.Neurons.ToList()[1] as IOutputNeuron;
// sigmoid function : (1.0 / (1 + Math.Exp(-1.0 * value)));
var h1Activation = ApplyActivationFunction((.75 * input1Hidden1Conn.Weight) + hidden1.Bias);
var h2Activation = ApplyActivationFunction((.75 * input1Hidden2Conn.Weight) + hidden2.Bias);
var o1Activation = ApplyActivationFunction((h1Activation * hidden1Output1Conn.Weight) + (h2Activation * hidden2Output1Conn.Weight) + output1.Bias);
var o2Activation = ApplyActivationFunction((h1Activation * hidden1Output2Conn.Weight) + (h2Activation * hidden2Output2Conn.Weight) + output2.Bias);
// Assert
Assert.IsTrue(o1Activation == output1After.ActivationLevel);
Assert.IsTrue(o2Activation == output2After.ActivationLevel);
}
public void DFFNeuralNetwork_WithLayers_InitializesLayers()
{
// Arrange
var layers = new List <INetworkLayer>()
{
_inputLayer, _hiddenLayer, _outputLayer
};
// Act
var network = new DFFNeuralNetwork(layers);
// Assert
Assert.IsNotNull(network.Layers);
Assert.IsTrue(network.Layers.Count() == 3);
Assert.IsTrue(network.Layers.Contains(_inputLayer));
Assert.IsTrue(network.Layers.Contains(_hiddenLayer));
Assert.IsTrue(network.Layers.Contains(_outputLayer));
}
/// <summary>
/// Creates and returns a neural network using configuration settings in the <paramref name="networkConfiguration"/> and the <paramref name="trainingEntry"/>.
/// </summary>
/// <param name="networkConfiguration"></param>
/// <param name="trainingDatasetEntries"></param>
/// <returns>Returns a neural network using configuration settings in the <paramref name="networkConfiguration"/> and the <paramref name="trainingEntry"/>.</returns>
private IDFFNeuralNetwork CreateNetwork(N networkConfiguration, INetworkTrainingIteration trainingEntry)
{
if (trainingEntry == null || trainingEntry.Inputs == null || trainingEntry.Outputs == null)
{
throw new ArgumentNullException("trainingEntry");
}
var numInputs = trainingEntry.Inputs.Count();
var numOutputs = trainingEntry.Outputs.Count();
var numHiddenLayers = networkConfiguration.NumHiddenLayers;
var numHiddenLayerNeurons = networkConfiguration.NumHiddenLayerNeurons;
// Setup and randomize network.
var network = new DFFNeuralNetwork(numInputs, numHiddenLayers, numHiddenLayerNeurons, numOutputs);
network.RandomizeNetwork();
return(network);
}
