public void TrainingsCycle(double[] echtBild, double[] animie)
{
encoder.SetInput(echtBild);
encoder.OutputBerechnen();
decoder.SetInput(encoder.GetOutput());
decoder.OutputBerechnen();
double[] encoderOutput = encoder.GetOutput();
encoder.SetInput(decoder.GetOutput());
encoder.OutputBerechnen();
encoder.DeltawertBerechen(encoderOutput);
discriminator.SetInput(decoder.GetOutput());
discriminator.OutputBerechnen();
discriminator.DeltawertBerechen(new double[] { 1 });
decoder.IsertDelta(discriminator.GetFirstLayerDelta());
decoder.Backpropagation();
encoder.Backpropagation();
discriminator.DeltawertBerechen(new double[] { 0 });
discriminator.Backpropagation();
discriminator.SetInput(animie);
discriminator.OutputBerechnen();
discriminator.DeltawertBerechen(new double[] { 1 });
discriminator.Backpropagation();
}
public void TrainDisciminator(int cycles, string animieFolder)
{
string[] files = Directory.GetFiles(animieFolder);
for (int i = 0; i < cycles; ++i)
{
discriminator.SetInput(MatrixMafs.ArrDivNum(GetDoubleArr(File.ReadAllBytes(files[i % files.Length])), 255));
discriminator.OutputBerechnen();
discriminator.DeltawertBerechen(new double[] { 1 });
discriminator.Backpropagation();
discriminator.SetInput(SimpleNet.GetRandArr(750000, 0, 1));
discriminator.OutputBerechnen();
discriminator.DeltawertBerechen(new double[] { 0 });
discriminator.Backpropagation();
if (i % 1000 == 0)
{
Console.WriteLine("Trainingscycle nr " + i);
}
}
}