public void PropagateLayer(BNet net, BLayer upper)
{
double sum;
for (int i = 1; i <= upper.Units; i++)
{
sum = 0;
for (int j = 0; j <= Units; j++)
{
sum += upper.Weights[i][j] * Output[j];
}
upper.Output[i] = 1 / (1 + Math.Exp(-net.Gain * sum));//sigmoid
}
}
public void BackpropagateLayer(BNet net, BLayer lower)
{
double Out, Err;
for (int i = 1; i <= lower.Units; i++)
{
Out = lower.Output[i];
Err = 0;
for (int j = 1; j <= Units; j++)
{
Err += Weights[j][i] * Error[j];
}
lower.Error[i] = net.Gain * Out * (1 - Out) * Err;
}
}
public BNet(int[] neurons, double gain, double eta, double alpha /*, int inputsCount, int layersCount*/) : base(neurons[0], neurons.Length)
{
this.inputsCount = neurons[0];
this.layersCount = neurons.Length;
this.Gain = gain;
this.Eta = eta;
this.Alpha = alpha;
this.Neu = neurons;
layers[0] = new BLayer(neurons[0], neurons[0], -1, true);
for (int l = 1; l < layersCount; l++)
{
layers[l] = new BLayer(neurons[l], neurons[l], neurons[l - 1], false);
}
in_layer = (BLayer)layers[0];
out_layer = (BLayer)layers[layersCount - 1];
}
