public void RunTraining(ref IdentityNeuron neuron, ref NeuronTrainingPattern[] trainingSet, int inputNumber, double trainingStep, int epochs)
{
if (neuron == null || trainingSet == null || trainingSet.Count() < 0 || epochs < 1)
{
System.Diagnostics.Debug.Write("Warning: Attempting to run training without all parameters correctly set.");
return;
}
int trainingNumber = trainingSet.Count();
resultWeights = null;
resultErrors = null;
resultWeights = new double[epochs][];
for (int i = 0; i < epochs; i++)
{
resultWeights[i] = new double[inputNumber];
}
resultErrors = new double[epochs];
for (int i = 0; i < epochs; i++)
{
double error = runTrainingIteration(ref neuron, ref trainingSet, trainingStep, trainingNumber, inputNumber);
resultErrors[i] = error;
resultWeights[i] = neuron.Weights;
//System.Diagnostics.Debug.Write("Weight after: " + i.ToString() + " epoch:" + resultWeights[i][0].ToString() + "\n");
System.Diagnostics.Debug.Write("Errors after: " + i.ToString() + " epoch: " + resultErrors[i].ToString() + "\n");
}
}
private double runTrainingIteration(ref IdentityNeuron neuron, ref NeuronTrainingPattern[] trainingSet, double trainingStep, int trainingNumber, int inputNumber)
{
double finalError = 0.0;
for (int i = 0; i < trainingNumber; i++)
{
double[] inputVectors = trainingSet[i].neuronInputVectors;
double computedOutputVector = neuron.Compute(inputVectors);
double error = trainingSet[i].neuronOutputVector - computedOutputVector;
for (int j = 0; j < inputNumber; j++)
{
neuron.Weights[j] = neuron.Weights[j] + trainingStep * error * inputVectors[j];
}
finalError += (error * error);
}
finalError /= trainingNumber;
return finalError;
}