public double LearnOnReducedData(int num_TrainingPerClass)
{
double meanSquareError = threshold + 1;
while (meanSquareError > threshold)
{
meanSquareError = 0;
for (int i = 0; i < num_Classes; i++)
{
for (int j = 1; j <= num_TrainingPerClass; j++)
{
String pattern = "img/" + classes[i] + "/Training/" + j + ".txt";
string[] lines = System.IO.File.ReadAllLines(pattern);
double[] input = HelperFunctions.ToDoubleArray(lines);
HelperFunctions.Normalize(ref input);
double[] output = new double[num_Classes];
for (int k = 0; k < num_Classes; k++)
{
if (k != i)
{
output[k] = 0;
}
else
{
output[k] = 1;
}
}
MLP.BackPropagate(PCA.ForwardPropagate(input), output);
}
}
for (int i = 0; i < num_Classes; i++)
{
for (int j = 1; j <= num_TrainingPerClass; j++)
{
String pattern = "img/" + classes[i] + "/Training/" + j + ".txt";
string[] lines = System.IO.File.ReadAllLines(pattern);
double[] input = HelperFunctions.ToDoubleArray(lines);
HelperFunctions.Normalize(ref input);
double[] output = MLP.ForwardPropagate(PCA.ForwardPropagate(input));
double[] desiredOutput = new double[num_Classes];
double error = 0;
for (int k = 0; k < num_Classes; k++)
{
if (k != i)
{
desiredOutput[k] = 0;
}
else
{
desiredOutput[k] = 1;
}
error = desiredOutput[k] - output[k];
meanSquareError += error * error;
}
}
}
meanSquareError /= num_TrainingPerClass * num_Classes * 2;
Console.WriteLine(meanSquareError);
}
return(meanSquareError);
}
