private void SigmoidActivation(DataSample sample)
{
if (sample.data.Length != inputSize)
{
Console.WriteLine("NOT the same size - " + sample.data.Length + " / " + inputSize);
return;
}
inputLayerValue = sample.data;
// First Layer Activation
for (int i = 0; i < hiddenSize; i++)
{
float sum = 0;
for (int j = 0; j < inputSize; j++)
{
sum += (float)(inputLayerValue[j] * inputLayerWeights[j, i]);
}
hiddenLayerValue[0, i] = ActivationFunctions.Sigmoid(sum);
}
// Hidden Layer Activation
for (int i = 1; i < hiddenDimension; i++)
{
for (int j = 0; j < hiddenSize; j++)
{
float sum = 0;
for (int k = 0; k < hiddenSize; k++)
{
sum += (hiddenLayerValue[i - 1, k] * hiddenLayerWeights[i - 1, j, k]);
}
hiddenLayerValue[i, j] = ActivationFunctions.Sigmoid(sum);
}
}
// Output Layer Activation
for (int i = 0; i < outputSize; i++)
{
float sum = 0;
for (int j = 0; j < hiddenSize; j++)
{
sum += (hiddenLayerValue[hiddenDimension - 1, j] * outputLayerWeights[i, j]);
}
outputLayerValue[i] = ActivationFunctions.Sigmoid(sum);
}
}
public void DataClassification(DataSample sample)
{
SigmoidActivation(sample);
if (ReturnCost(CreateIdealVector(sample.label)) < .5)
{
Console.Write("Classified: ");
float maxVal = outputLayerValue[0];
int index = 0;
for (int i = 0; i < outputSize; i++)
{
if (outputLayerValue[i] > maxVal)
{
maxVal = outputLayerValue[i];
index = i;
}
}
switch (index)
{
case 0:
Console.Write("Classified: Blank");
break;
case 1:
Console.Write("Classified: Noise");
break;
case 2:
Console.Write("Classified: Hej");
break;
case 3:
Console.Write("Classified: Siri");
break;
}
}
}
static void NetworkTraining(int inputSize, int width, int height, int itterations)
{
int outputSize = DataSample.labelSize;
string databasePath1 = @"C:/Users/Mikkel/Documents/MTA18434_SemesterProject/ML_Sound_Samples/Assets/Resources/SampleDatabase/Database.json";
string networkPath1 = Path.GetDirectoryName(Assembly.GetEntryAssembly().Location) + "/NetworkSave01.json";
if (File.Exists(networkPath1))
{
network = NeuralNetwork.LoadNetwork(networkPath1);
if (network.inputSize != inputSize)
{
throw new Exception();
}
if (network.hiddenDimension != width)
{
throw new Exception();
}
if (network.hiddenSize != height)
{
throw new Exception();
}
}
else
{
Console.WriteLine("Initializing NN");
network = new NeuralNetwork(inputSize, width, height, outputSize);
Console.WriteLine("NN Initialized");
}
if (File.Exists(databasePath1))
{
Console.WriteLine(File.Exists(databasePath1) ? "Database exists." : "File does not exist.");
SampleDatabase temp = null;
using (StreamReader r = new StreamReader("C:/Users/Mikkel/Documents/MTA18434_SemesterProject/ML_Sound_Samples/Assets/Resources/SampleDatabase/Database.json"))
{
using (JsonReader reader = new JsonTextReader(r))
{
JsonSerializer serializer = new JsonSerializer();
Console.WriteLine("Deserializing");
temp = serializer.Deserialize <SampleDatabase>(reader);
}
}
Console.WriteLine(temp.database[0].data.Length);
DataSample[] trainingSamples = new DataSample[10];
Random rand = new Random();
for (int i = 0; i < itterations; i++)
{
Console.WriteLine("Progress: " + i + " / " + itterations);
// pick 10 samples
for (int j = 0; j < 10; j++)
{
int num = rand.Next(0, temp.database.Length);
trainingSamples[j] = new DataSample(temp.database[num].data, temp.database[num].label);
//Console.WriteLine("Database sample " + temp.database[num].data[0] + " " + temp.database[num].label);
}
network.TrainNetwork(trainingSamples);
}
Console.WriteLine("Saving network");
network.SaveNetwork(networkPath1);
Console.WriteLine("Network saved");
}
else
{
Console.WriteLine(File.Exists(databasePath1) ? "File exists." : "File does not exist.");
throw new Exception();
}
}
private GradientWeightVector GradientBackProp(DataSample sample)
{
// Backprop Output Layer
GradientWeightVector weightVector = new GradientWeightVector(this);
float[] idealVector = CreateIdealVector(sample.label);
float[] outputLayerDelta = new float[outputSize];
float[,] hiddenLayerDelta = new float[hiddenDimension - 1, hiddenSize];
float[] inputLayerDelta = new float[inputSize];
SigmoidActivation(sample);
Console.WriteLine("Label: " + sample.label);
Console.WriteLine();
PrintOutput();
Console.WriteLine();
Console.WriteLine("Cost after Activation is: " + ReturnCost(CreateIdealVector(sample.label)));
PrintAverageCost();
// Backpropagation for the outputlayer
for (int i = 0; i < outputSize; i++)
{
outputLayerDelta[i] = -(idealVector[i] - outputLayerValue[i]) * outputLayerValue[i] *
(1 - outputLayerValue[i]);
for (int j = 0; j < hiddenSize; j++)
{
weightVector.outputLayerWeights[i, j] = learningRate * (outputLayerDelta[i] * hiddenLayerValue[hiddenDimension - 1, j]);
}
}
// Finding delta and gradients for the last hidden layer
for (int i = 0; i < hiddenSize; i++)
{
float sum = 0;
for (int j = 0; j < outputSize; j++)
{
for (int k = 0; k < hiddenSize; k++)
{
sum += outputLayerDelta[j] * outputLayerWeights[j, k];
}
}
hiddenLayerDelta[hiddenDimension - 2, i] = sum;
for (int j = 0; j < hiddenSize; j++)
{
weightVector.hiddenLayerWeights[hiddenDimension - 2, i, j] = learningRate * (hiddenLayerDelta[hiddenDimension - 2, i] *
hiddenLayerValue[hiddenDimension - 1, i] * (1 - hiddenLayerValue[hiddenDimension - 1, i]) *
hiddenLayerValue[hiddenDimension - 2, j]);
}
}
for (int i = hiddenDimension - 2; i > 0; i--)
{
for (int j = 0; j < hiddenSize; j++)
{
float sum = 0;
for (int k = 0; k < hiddenSize; k++)
{
sum += hiddenLayerDelta[i, j] * hiddenLayerWeights[i, j, k];
}
hiddenLayerDelta[i, j] = sum;
for (int k = 0; k < hiddenSize; k++)
{
weightVector.hiddenLayerWeights[i, j, k] = learningRate * (hiddenLayerDelta[i, j] * hiddenLayerValue[i, j] *
(1 - hiddenLayerValue[i, j]) * hiddenLayerValue[i - 1, j]);
}
}
}
// BackPropagation for the input layer
for (int i = 0; i < inputSize; i++)
{
float sum = 0;
for (int j = 0; j < hiddenSize; j++)
{
sum += hiddenLayerDelta[0, j] * inputLayerWeights[i, j];
}
inputLayerDelta[i] = sum;
for (int j = 0; j < hiddenSize; j++)
{
weightVector.inputLayerWeights[i, j] = learningRate * (inputLayerDelta[i] * hiddenLayerValue[0, j] *
(1 - hiddenLayerValue[0, j]) * (float)inputLayerValue[i]);
}
}
return(weightVector);
}
