private void DoBackPropagation()
{
for (int i = 0; i < chromosoms.Count; i++)
{
ListDataSet lds = new ListDataSet(listDataSet);
for (int j = 0; j < lds.Count; j++)
{
int popCount = 0;
for (int k = 0; k < chromosoms[i].Length; k++)
{
if (chromosoms[i][k] == 0)
{
lds[j].RemoveBit(k - popCount);
popCount++;
}
}
}
NeuralNetwork nn = new NeuralNetwork();
nn.InitialiseNetwork(lds[0].AttributeCount, lds[0].AttributeCount / 2, classificationClass.TargetCount);
nn.InitialiseWeight();
BackPropagation bp = new BackPropagation();
bp.Initialise(nn, lds, classificationClass);
bp.Run();
FeedForward ff = new FeedForward();
ff.Initialise(nn, lds);
int totalCorrect = 0;
for (int j = 0; j < lds.Count; j++)
{
ff.Run(j);
//nn.Print();
//foreach (int ac in ff.GetActualClass())
// Console.Write(ac + " ");
//Console.Write(" --> ");
//foreach (int ac in classificationClass.GetTarget(lds[j].ClassName))
// Console.Write(ac + " ");
//Console.WriteLine();
bool correct = true;
int[] targetClass = classificationClass.GetTarget(lds[j].ClassName);
for (int k = 0; k < ff.GetActualClass().Length; k++)
{
if (targetClass[k] != ff.GetActualClass()[k])
{
correct = false;
}
}
if (correct)
{
totalCorrect++;
}
}
//Console.WriteLine("total: " + totalCorrect + "/" + lds.Count);
chromosoms[i].FitnessValue = totalCorrect / (float)lds.Count;
//chromosoms[i].Print();
}
}
public void Initialise(NeuralNetwork nn, ListDataSet dsl, ClassificationClass cc)
{
feedForward = new FeedForward();
feedForward.Initialise(nn, dsl);
classificationClass = cc;
}
public void Initialise(NeuralNetwork nn, ListDataSet dsl)
{
neuralNetwork = nn;
dataSetList = dsl;
}
