//Back Propagate for Hidden Neuron
public int BackPropag(main.Network net, int mPos, double learnRate)
{
this.dE_over_dOut = 0;
if (this.actFunc == actFuncType.logistic)
{
for (int i = 0; i < net.numNeuronsEachLayer[layer + 1]; i++)
{
//if (net[i, this.layer + 1].weights[mPos] != 0)
//{
Neuron outNeuron = net.nNarray[i, this.layer + 1];
this.dE_over_dOut += outNeuron.dE_over_dOut * outNeuron.dOut_over_dNet * outNeuron.weights[mPos];
//}
}
this.dOut_over_dNet = this.postActOutput * (1 - this.postActOutput);
for (int i = 0; i < this.inputs.Count; i++)
{
this.dNet_over_dW.Add(this.inputs[i]);
this.dE_over_dW.Add(this.dE_over_dOut * this.dOut_over_dNet * this.dNet_over_dW[i]);
this.newWeights.Add(this.weights[i] - learnRate * this.dE_over_dW[i]);
}
return(0);
}
else
{
//return -1;
throw new Exception("Activation function not ready for back propagation.");
}
}
//Minimal Hidden or Output Neuron Initializing Constructor
public Neuron(main.Network inNet, List <List <int> > inInputNs, int inLayer)
{
this.inputNeurons = inInputNs;
this.layer = inLayer;
this.weights = new List <double>();
for (int i = 0; i < inNet.numNeuronsEachLayer[inLayer - 1]; i++)
{
if (inNet.nNarray[i, inLayer - 1] != null)
{
this.weights.Add(DEFAULT_WEIGHT);
}
else
{
this.weights.Add(0f);
}
}
}
