public void AddBias()
{
Neuron lastNeuron = Neurons.Last();
Neuron newNeuron = new Neuron(
lastNeuron.Dendrites.Count(),
lastNeuron.ID + 1);
newNeuron.Value = 1;
newNeuron.Bias = true;
Neurons.Add(newNeuron);
}
/// <summary>
/// Perceptron is trained with the help of a teacher using the method of back propagation of an error using deltas
/// </summary>
/// <param name="ideal">The correct output value</param>
public (Vector output, Vector error) Learn(Vector input, Vector ideal)
{
CheckIdealVector(ideal, Neurons.Last());
var actual = Run(input);
var error = new Vector(actual.Length);
for (var i = 0; i < ideal.Length; i++)
{
ideal[i] = Activation.Func(ideal[i]);
error[i] = GetError(actual[i], ideal[i]);
}
LearnWithBackPropagationError(actual, ideal);
SetNextEpoch();
return(actual, error);
}
/// <summary>
/// Calculates the output vector of the neural network
/// </summary>
public Vector Run(Vector input)
{
InitializeNeuronsWithInput(input);
for (var i = 0; i < Neurons.Length - 1; i++)
{
for (int j = 0; j < Neurons[i + 1].Length; j++)
{
for (int k = 0; k < Neurons[i].Length; k++)
{
Neurons[i + 1][j] += Weights[i][j][k] * Neurons[i][k];
}
Neurons[i + 1][j] = Activation.Func(Neurons[i + 1][j]);
}
}
return(Neurons.Last());
}
private void AddWeights(Layer prevLayer, int index)
{
List <Neuron> lastNeurons = new List <Neuron>();
List <double> weights = new List <double>();
List <double> inputs = new List <double>();
for (int i = 0; i < NumOfPrevious; i++)
{
for (int j = 0; j < NumOfPrevious; j++)
{
weights.Add(0.5);
inputs.Add(prevLayer.Neurons.Find(x => x.MapX == i && x.MapY == j).Output);
lastNeurons.Add(prevLayer.Neurons.Find(x => x.MapX == i && x.MapY == j));
}
}
Neurons.Add(new Neuron(inputs, weights, _NeuronType, 0, index));
Neurons.Last().LastNeurons = lastNeurons;
}
protected void OverlayMatrix(int shiftI, int shiftJ, Layer prevLayer, int step)
{
int wi = 0; int wj = 0;
List <Neuron> lastNeurons = new List <Neuron>();
List <double> weights = new List <double>();
List <double> inputs = new List <double>();
for (int i = 0 + shiftI; i < Weights.Length + shiftI; i++)
{
for (int j = 0 + shiftJ; j < Weights.Length + shiftJ; j++)
{
weights.Add(Weights[wi][wj]);
inputs.Add(prevLayer.Neurons.Find(x => x.MapX == i && x.MapY == j).Output);
lastNeurons.Add(prevLayer.Neurons.Find(x => x.MapX == i && x.MapY == j));
wj++;
}
wj = 0; wi++;
}
Neurons.Add(new Neuron(inputs, weights, _NeuronType, shiftI / step, shiftJ / step));
Neurons.Last().LastNeurons = lastNeurons;
}
public T[] Calc()
{
return(Neurons.Last().Select(n => n.Calc()).ToArray());
}
