public void preparePointList(TrainingResultP trainingResult, TrainingSetings.ActiviationFunction activationFunction)
{
float x1 = 0.01f;
float x2 = 0.01f;
for (int x = 0; x <= 100; x++)
{
x2 = 0.01f;
for (int y = 0; y <= 100; y++)
{
PointInfo point = preparePoint(trainingResult, x1, x2, activationFunction);
if (point.color == 0.00f)
{
list0.Add(point);
}
else
{
list1.Add(point);
}
x2 += 0.01f;
}
x1 += 0.01f;
}
}
private void TrainingStartButton_Click(object sender, EventArgs e)
{
Debug.WriteLine("Click");
try
{
trainingSetings.learningRate = float.Parse(LearningRateTB.Text, CultureInfo.InvariantCulture);
trainingSetings.maxError = float.Parse(MaxErrorTB.Text, CultureInfo.InvariantCulture);
Debug.WriteLine("Jestem w try");
if (trainingSetings.gateType == TrainingSetings.GateType.XOR)
{
Debug.WriteLine("Xor training in Form1");
xorTrResult = trainingService.trainXOR(trainingSetings, gateTrValuesContainer);
showTrainingRaport(xorTrResult.raport);
}
else
{
trainingResult = trainingService.train(trainingSetings, gateTrValuesContainer);
showTrainingRaport(trainingResult.raport);
}
}
catch (NotImplementedException exception)
{
MessageBox.Show(exception.Message);
}
catch (ArgumentException exception)
{
MessageBox.Show(exception.Message);
}
catch (FormatException exception)
{
MessageBox.Show(exception.Message);
}
}
public void showChart(TrainingResultP trainingResult, TrainingSetings trainSettings)
{
list0 = new List <PointInfo>();
list1 = new List <PointInfo>();
activatonFunctions.maxError = trainSettings.maxError;
preparePointList(trainingResult, trainSettings.activiationFunction);
float[,] table0 = changeListToArray(list0);
float[,] table1 = changeListToArray(list1);
ChartForm chartForm = new ChartForm(table0, table1);
chartForm.Show();
}
public PointInfo preparePoint(TrainingResultP trainingResult, float x, float y, TrainingSetings.ActiviationFunction activiationFunction)
{
float e = x * trainingResult.w1 + y * trainingResult.w1 + trainingResult.biasI;
float result = 0.00f;
if (activiationFunction == TrainingSetings.ActiviationFunction.JUMP)
{
result = activatonFunctions.jumpActivationFuntion(e);
}
else
{
result = activatonFunctions.sigmoidActivationFunction(e);
result = activatonFunctions.correctionForSigmoid(result);
}
return(new PointInfo(x, y, result));
}
private void trainByOneNeuron(ActivatonFunctions.Del activationFunction)
{
if (trainingSetings.gateType == TrainingSetings.GateType.XOR)
{
throw new NotImplementedException("Brak możliwości trenowanie XOR jednym nuronem");
}
tResult = new TrainingResultP();
firstLineBuilder.Append(" pojedynczy neuron.");
tResult.raport.Add(firstLineBuilder.ToString());
bool isWeightChanged = true;
Random random = new Random();
float w1 = (float)random.NextDouble();
float w2 = (float)random.NextDouble();
float bias = 1.00f;
float e = 0.00f;
float y = 0.00f;
tResult.raport.Add("Początkowe w1: " + w1);
tResult.raport.Add("Początkowe w2: " + w2);
tResult.raport.Add("Początkowy bias: " + bias);
List <float> desireResults = valuesContainer.results[trainingSetings.gateType.ToString()];
if (desireResults == null)
{
Debug.WriteLine("DesireResults jest nullem");
}
else
{
Debug.WriteLine("desireResults");
foreach (float result in desireResults)
{
Debug.WriteLine(result);
}
}
Debug.WriteLine("Typ bramki do szkolenia: " + trainingSetings.gateType);
Debug.WriteLine("Funkcja aktywacyjna: " + trainingSetings.activiationFunction);
int epochs = 0;
while (isWeightChanged)
{
isWeightChanged = false;
epochs++;
for (int i = 0; i < valuesContainer.x1Values.Count; i++)
{
e = w1 * valuesContainer.x1Values[i] + w2 * valuesContainer.x2Values[i] + bias;
y = activationFunction(e);
y = activationFunctions.correctionForSigmoid(y);
if (y != desireResults[i])
{
tResult.raport.Add("");
tResult.raport.Add("desire = " + desireResults[i] + ", y = " + y + ", e: " + e);
tResult.raport.Add("Zmiana w1, w2, bias");
w1 = w1 + trainingSetings.learningRate * (desireResults[i] - y) * valuesContainer.x1Values[i];
w2 = w2 + trainingSetings.learningRate * (desireResults[i] - y) * valuesContainer.x2Values[i];
bias = bias + trainingSetings.learningRate * (desireResults[i] - y);
tResult.raport.Add("w1: " + w1);
tResult.raport.Add("w2: " + w2);
tResult.raport.Add("bias: " + bias);
isWeightChanged = true;
}
}
}
tResult.w1 = w1;
tResult.w2 = w2;
tResult.biasI = bias;
tResult.raport.Add("Finalne w1: " + w1);
tResult.raport.Add("Finalene w2: " + w2);
tResult.raport.Add("Finalne bias: " + bias);
tResult.raport.Add("Liczba epok: " + epochs);
}
