/// <summary>
/// Connects current layer to output (right layer)
/// It should not connect biases with previous layer
/// </summary>
/// <param name="outputNeuronLayer">Layer from the right side</param>
/// <param name="max">Max value of weight</param>
/// <param name="min">Min value of weight</param>
public void ConnectToOutputLayerWithBiases(NeuronLayer outputNeuronLayer, double max, double min)
{
var random = new Random();
IEnumerable <Neuron> outputNeurons;
//handle case, when we have one neuron on the output
if (outputNeuronLayer.Neurons.Count > 1)
{
outputNeurons = outputNeuronLayer.Neurons.Take(outputNeuronLayer.Neurons.Count - 1);
}
else
{
outputNeurons = outputNeuronLayer.Neurons;
}
//leave the bias
foreach (var outputNeuron in outputNeurons)
{
//connect all neurons with bias from current layer to right layer
foreach (var inputNeuron in Neurons)
{
var value = random.NextDouble() * (max - min) + min;
inputNeuron.ConnectToOutputNeuron(
outputNeuron, (float)value);
}
}
}
/// <summary>
/// Connects current neuron layer with right (next) layer
/// It will connect all neurons of current layer and all neurons of right layer
/// </summary>
/// <param name="outputNeuronLayer">Layer from right side (next)</param>
public void ConnectToOutputLayer(NeuronLayer outputNeuronLayer)
{
var random = new Random();
foreach (var localNeuron in Neurons)
{
foreach (var outputNeuron in outputNeuronLayer.Neurons)
{
localNeuron.ConnectToOutputNeuron(outputNeuron, (float)random.NextDouble());
}
}
}
