static public double[] LayerDerivatives(out ConectionsInfo conD, double[] foreLayerD, ConectionsInfo con, double[] backLayerA)
{
var neuronD = new double[con.Back];
var weightD = new double[con.Fore, con.Back];
var biasD = new double[con.Fore];
var Z = GetZ(backLayerA, con);
for (int j = 0; j < con.Fore; j++)
{
for (int k = 0; k < con.Back; k++)
{
weightD[j, k] = backLayerA[k] * DSigmoid(Z[j]) * foreLayerD[j];
}
biasD[j] = DSigmoid(Z[j]) * foreLayerD[j];
}
for (int k = 0; k < con.Back; k++)
{
neuronD[k] = 0;
for (int j = 0; j < con.Fore; j++)
{
neuronD[k] += con.WeightMatrix[j, k] * DSigmoid(Z[j]) * foreLayerD[j];
}
}
conD = new ConectionsInfo(con.Fore, con.Back, weightD, biasD);
return(neuronD);
}
static public double[] CountALayer(double[] back, ConectionsInfo conections)
{
var neuronStates = GetZ(back, conections);
for (int i = 0; i < conections.Fore; i++)
{
neuronStates[i] = Sigmoid(neuronStates[i]);
}
return(neuronStates);
}
static double[] GetZ(double[] back, ConectionsInfo conections)
{
var Z = new double[conections.Fore];
for (int i = 0; i < conections.Fore; i++)
{
Z[i] = conections.BiasVector[i];
for (int j = 0; j < conections.Back; j++)
{
Z[i] += conections.WeightMatrix[i, j] * back[j];
}
}
return(Z);
}
