private XorTrResult trainByTwoNeuronsBySigmoid()
{
Debug.WriteLine("Działa");
float[,] inputs =
{
{ 0, 0 },
{ 0, 1 },
{ 1, 0 },
{ 1, 1 }
};
float[] results = { 0, 1, 1, 0 };
Neuron hiddenNeuron1 = new Neuron();
Neuron hiddenNeuron2 = new Neuron();
Neuron outputNeuron = new Neuron();
List <String> raport = new List <String>();
hiddenNeuron1.randomizeWeights();
raport.Add("Inicjalizacja: ");
raport.Add("hiddenNeuron1.weights[0]: " + hiddenNeuron1.weights[0]);
raport.Add("hiddenNeuron1.weights[1]: " + hiddenNeuron1.weights[1]);
hiddenNeuron2.randomizeWeights();
raport.Add("hiddenNeuron2.weights[0]: " + hiddenNeuron2.weights[0]);
raport.Add("hiddenNeuron2.weights[1]: " + hiddenNeuron2.weights[1]);
outputNeuron.randomizeWeights();
raport.Add("outputNeuron.weights[0]: " + outputNeuron.weights[0]);
raport.Add("outputNeuron.weights[1]: " + outputNeuron.weights[1]);
bool isChanged = true;
int epochs = 0;
while (isChanged)
{
isChanged = false;
for (int i = 0; i < 4; i++)
{
// 1) forward propagation (calculates output)
hiddenNeuron1.inputs = new float[] { inputs[i, 0], inputs[i, 1] };
hiddenNeuron2.inputs = new float[] { inputs[i, 0], inputs[i, 1] };
outputNeuron.inputs = new float[] { hiddenNeuron1.output(), hiddenNeuron2.output() };
float result = outputNeuron.output();
Debug.WriteLine("{0} xor {1} = {2}", inputs[i, 0], inputs[i, 1], result);
// 2) back propagation (adjusts weights)
// adjusts the weight of the output neuron, based on its error
result = activationFunctions.correctionForSigmoid(result);
Debug.WriteLine("output po korekcji: " + result);
if (result != results[i])
{
isChanged = true;
outputNeuron.error = Sigmoid.derivative(outputNeuron.output()) * (results[i] - outputNeuron.output());
outputNeuron.adjustWeights(trainingSetings.learningRate);
// then adjusts the hidden neurons' weights, based on their errors
hiddenNeuron1.error = Sigmoid.derivative(hiddenNeuron1.output()) * outputNeuron.error * outputNeuron.weights[0];
hiddenNeuron2.error = Sigmoid.derivative(hiddenNeuron2.output()) * outputNeuron.error * outputNeuron.weights[1];
hiddenNeuron1.adjustWeights(trainingSetings.learningRate);
hiddenNeuron2.adjustWeights(trainingSetings.learningRate);
raport.Add("Zmiana wag: ");
raport.Add("hiddenNeuron1.weights[0]: " + hiddenNeuron1.weights[0]);
raport.Add("hiddenNeuron1.weights[1]: " + hiddenNeuron1.weights[1]);
raport.Add("hiddenNeuron2.weights[0]: " + hiddenNeuron2.weights[0]);
raport.Add("hiddenNeuron2.weights[1]: " + hiddenNeuron2.weights[1]);
raport.Add("outputNeuron.weights[0]: " + outputNeuron.weights[0]);
raport.Add("outputNeuron.weights[1]: " + outputNeuron.weights[1]);
}
}
epochs++;
}
raport.Add("Ilość epochs: " + epochs);
return(new XorTrResult(hiddenNeuron1, hiddenNeuron2, outputNeuron, raport));
}