internal void Apply(NeuralTrainLayer neuralTrainLayer, float learningRate)
{
if (neuralTrainLayer.TrainingCount == 0)
{
return;
}
float multiplyer = learningRate / neuralTrainLayer.TrainingCount;
float[] bShift = neuralTrainLayer.DCDB.Multiply(multiplyer);
float[] wShift = neuralTrainLayer.DCDW.Multiply(multiplyer);
Bias = Bias.Minus(bShift);
Weight = Weight.Minus(wShift);
}
public void UpdateWeights(Matrix targets, double learningRate)
{
if (Next == null)
{
Errors = Matrix.Subtract(targets, Outputs);
Matrix gradient = Matrix.Copy(Outputs);
gradient.Map(x => ActivationFc_derivitiv(x));
gradient.Multiply(Errors);
gradient.Multiply(learningRate);
// Matrix Weight_T = Matrix.Transpose(Inputs);
Matrix delta = Matrix.Multiply(gradient, Matrix.Transpose(Inputs));
Weights.Add(delta);
Bias.Add(gradient);
}
else
{
Errors = Matrix.Multiply(Matrix.Transpose(Next.Weights), Next.Errors);
Matrix gradient = Matrix.Copy(Outputs);
gradient.Map(x => ActivationFc_derivitiv(x));
gradient.Multiply(Errors);
gradient.Multiply(learningRate);
Matrix delta = Matrix.Multiply(gradient, Matrix.Transpose(Inputs));
Weights.Add(delta);
Bias.Add(gradient);
}
if (Prev != null)
{
Prev.UpdateWeights(targets, learningRate);
}
}
public OutputNeuron(ITransferFunction tFunc) : base(tFunc)
{
Bias = new Bias();
}
public sealed override void CalculateOutput()
{
Output += Bias.Calculate();
Output = base.TransferFunction.Evaluate(Output);
}
