protected Layer(int neuronsNumber, IActivation activationFunction, INetwork networkType, Bias bias = null) : base(neuronsNumber)
{
this.networkType = networkType;
if (activationFunction is null)
{
this.activationFunction = new SigmoidFunction();
}
if (this is OutputLayer && networkType is RegressionNetwork)
{
this.activationFunction = new IdentityFunction();
}
else
{
this.activationFunction = activationFunction;
}
//if (!(activationFunction is null))
// this.activationFunction = activationFunction;
for (int i = 0; i < neuronsNumber; i++)
{
this.neurons.Add(new Neuron(activationFunction, this.networkType));
}
if (bias == null)
{
return;
}
//neurons.Add(new InputNeuron(bias.Value));
foreach (var neuron in this.neurons)
{
((Neuron)neuron).AddPredecessors(new List <Tuple <INeuron, double> >()
{
new Tuple <INeuron, double>(new InputNeuron(bias.Value), bias.Wage)
});
}
}
//readonly List<Tuple<INeuron, double>> incomingLayerTuple;
public HiddenLayer(LayerBase incomingLayer, List <List <double> > initialWages, IActivation activationFunction, INetwork networkType, Bias bias = null)
: base(initialWages.Count, activationFunction, networkType, bias)
{
//this.networkType = networkType;
if (activationFunction is null)
{
this.activationFunction = new SigmoidFunction();
}
for (int i = 0; i < this.neurons.Count; i++)
{
var predecessors = new List <Tuple <INeuron, double> >();
for (int j = 0; j < incomingLayer.Neurons.Count; j++)
{
predecessors.Add(new Tuple <INeuron, double>(incomingLayer.Neurons[j], initialWages[i][j]));
}
this.GetNeuron(i).AddPredecessors(predecessors);
}
}
public OutputLayer(LayerBase incomingLayer, List <List <double> > initialWages, IActivation activationFunction, INetwork networkType, Bias bias = null)
: base(incomingLayer, initialWages, activationFunction, networkType, bias)
{
//if (activationFunction is null)
this.activationFunction = activationFunction;
this.networkType = networkType;
}
