public void CalculateOutput()
{
foreach (var inputNeuron in InputLayer.Neurons)
{
inputNeuron.Input = Activation.Sigmoid(inputNeuron.Input);
inputNeuron.Activate();
}
for (var i = 0; i < HiddenLayers.Count; i++)
{
foreach (var currentLayerNeuron in HiddenLayers[i].Neurons)
{
if (i == 0)
{
//Cross point between first hidden layer and input layer
foreach (var inputNeuron in InputLayer.Neurons)
{
currentLayerNeuron.Input += inputNeuron.Output;
}
}
else
{
foreach (var hiddenLayerNeuron in HiddenLayers[i - 1].Neurons)
{
currentLayerNeuron.Input += hiddenLayerNeuron.Output;
}
}
currentLayerNeuron.Activate();
}
}
foreach (var outputNeuron in OutputLayer.Neurons)
{
foreach (var hiddenLayerNeuron in HiddenLayers[HiddenLayers.Count - 1].Neurons)
{
outputNeuron.Input += hiddenLayerNeuron.Output;
}
outputNeuron.Activate();
}
}
public Dropout(Activation activation, double width = 0.25f) : base(activation)
{
this.width = width;
}
public Layer(int dimIn, int dimOut, Activation activation = null)
{
_W = new double[dimIn, dimOut];
_b = new double[dimOut];
_activation = activation;
}
public Layer(Activation activation)
{
func = activation;
}
public MaxPool2D(Activation activation, int width, int height) : base(activation)
{
this.width = width;
this.height = height;
}
public FullyConnLayar(Activation activation, Size output) : base(activation)
{
this.output_size = output;
}
public Conv2D(Activation activation, int width, int height, int count = 1) : base(activation)
{
this.width = width;
this.height = height;
this.count = count;
}
protected double GetChunkOfDeltaH(double weightSynapse, double neuronInBeginSynapse, double deltaInEndSynapse) => weightSynapse *deltaInEndSynapse *Activation.DeriveFunc(neuronInBeginSynapse);
protected Vector GetDelta0(Vector actual, Vector ideal) => Vector.Combine(actual, ideal, (actualItem, idealItem) => (idealItem - actualItem) * Activation.DeriveFunc(actualItem));