public Dendrite CreateDendrite()
{
Dendrite d = new Dendrite();
d.Id = Guid.NewGuid().ToString();
d.Weight = 0f;
d.NewWeight = null;
Dendrites.Add(d);
return(d);
}
public static NeuralNetwork Create(int[] layer_definition, bool include_bias_neurons)
{
NeuralNetwork nn = new NeuralNetwork();
List <Neuron> PreviousLayerNeurons = new List <Neuron>();
for (int t = 1; t <= layer_definition.Length; t++)
{
//Create this layers neurons
List <Neuron> ThisLayerNeurons = new List <Neuron>();
int NumberOfNeuronsForThisLayer = layer_definition[t - 1];
for (int nc = 1; nc <= NumberOfNeuronsForThisLayer; nc++)
{
Neuron n = nn.CreateNeuron();
ThisLayerNeurons.Add(n);
}
//If there are neurons in the "PreviousLayerNeurons" then connect the new ones to those
if (PreviousLayerNeurons.Count > 0)
{
foreach (Neuron tn in ThisLayerNeurons)
{
foreach (Neuron pn in PreviousLayerNeurons)
{
Dendrite d = nn.CreateDendrite();
d.InputNeuronId = pn.Id;
d.OutputNeuronId = tn.Id;
}
}
}
//Reset for next round
PreviousLayerNeurons.Clear();
PreviousLayerNeurons.AddRange(ThisLayerNeurons);
ThisLayerNeurons.Clear();
//Add a bias neuron if asked to (this has to be done after the full connections have been made in the previous step.
if (include_bias_neurons == true)
{
if (t != layer_definition.Length) //If it is not the last layer
{
Neuron bn = nn.CreateNeuron();
bn.InputValue = 1;
bn.IsBiasNeuron = true;
PreviousLayerNeurons.Add(bn);
}
}
}
//Randomize all weights
nn.RandomizeAllWeights();
return(nn);
}