/// DO POPRAWY
public static float Learn(Network network, DataSet data)
{
float meanError = 0.0f;
List <List <float> > placeForErrors = CreateNodesForErrors(network);
List <List <float> > placeForOutputs = CreateNodesForErrors(network);
for (int i = 0; i < data.Input.Count; i++)
{
float[] solution = new float [network.Classes];
solution = NetworkCalculation.CalculateSingleRecord(network, data.Input[i].ToArray(), placeForOutputs);
float[] MSError = SquaredError(solution, data.Output[i].ToArray());
float[] lastLayerError = LastLayerError(solution, data.Output[i].ToArray());
float tmpError = 0.0f;
for (int j = 0; j < MSError.Length; j++)
{
tmpError += MSError[j];
}
CalculateNodeErrors(placeForErrors, lastLayerError, network);
ModifyWages(network, placeForErrors, placeForOutputs, network.LearningFactor);
meanError += tmpError;
}
meanError = meanError / (float)data.Input.Count;
return(meanError);
}
public static float GetTestingSetError(Network network, DataSet testingSet)
{
float error = 0.0f;
int correctAnswerCounter = 0;
List <List <float> > placeForOutputs = LearningHelper.CreateNodesForErrors(network);
for (int i = 0; i < testingSet.Input.Count; i++)
{
float[] solution = new float[network.Classes];
solution = NetworkCalculation.CalculateSingleRecord(network, testingSet.Input[i].ToArray(), placeForOutputs);
bool areSame = CompareAnswers(solution, testingSet.Output[i].ToArray());
if (areSame)
{
correctAnswerCounter++;
}
}
error = (float)(((double)correctAnswerCounter * 100.0) / (double)testingSet.Output.Count);
return(error);
}