static double FitFunc(NeuralNetwork nn)
{
const double step = 0.01d;
const double start = -1;
const double end = 1;
var iterationCount = Math.Abs(start - end) / step;
double errorRate = 0d;
for (double i = start; i <= end; i += step)
{
errorRate += Math.Abs(nn.Solve(i) - func(i));
}
return 1d / (1 + errorRate / iterationCount);
}
public Species(Func <NeuralNetwork, double> fitnessFunc, NeuralNetwork network)
: this(fitnessFunc, network, null)
{
}
internal static IComputingNode GetRandomNode(this NeuralNetwork network)
{
return(network.Synapses[Rnd.Next(0, network.Synapses.Count)]._input);
}
internal static Neuron GetRandomNeuron(this NeuralNetwork network)
{
return(network.Neurons[Rnd.Next(0, network.Neurons.Count)]);
}
internal static Synapse GetRandomSynapse(this NeuralNetwork network)
{
return(network.Synapses[Rnd.Next(0, network.Synapses.Count)]);
}
public static NeuralNetwork ConnectMoreToOne(NeuralNetwork outNetwork, params NeuralNetwork[] networks)
{
/*
* //TODO: пофиксить Id нейронов
* if (networks.Sum(n => n.Effectors.Count) != outNetwork.Receptors.Count(r => typeof(Receptor)==r.GetType()))
* throw new Exception("Количество входов выходной сети должно равняться сумме всех выходов входных");
*
* if (networks.Max(n => n.Receptors.Count(r => typeof(Receptor) == r.GetType())) != networks.Min(n => n.Receptors.Count(r => typeof(Receptor) == r.GetType())))
* throw new Exception("Должно быть одинаковое число входов выходной сети (для объединения в один)");
*
* Receptor alphaReceptor = new Receptor(); //TODO: входных рецепторов может быть много
*
*
*
* List<Neuron> neuronsOnLatLayer = new List<Neuron>();
* var maxOfAllLayer = 0;
*
* List<Receptor> receptors = new List<Receptor>();
* List<Neuron> neurons = new List<Neuron>();
* List<Synapse> synapses = new List<Synapse>();
*
* foreach (var network in networks)
* {
* var maxLayer = network.Neurons.Max(n => n.LayerId);
* neuronsOnLatLayer.AddRange(network.Neurons.Where(n => n.LayerId == maxLayer).ToList());
*
* maxOfAllLayer = maxLayer > maxOfAllLayer ? maxLayer : maxOfAllLayer;
*
* foreach (var receptor in network.Receptors)
* {
* receptor.RecurrentCallback = alphaReceptor.Compute;
* receptor.Recurrent = true;
* }
*
* receptors.AddRange(network.Receptors);
* neurons.AddRange(network.Neurons);
* synapses.AddRange(network.Synapses);
* }
*
* for (int i = 0; i < outNetwork.ReceptorCount; i++)
* {
* var receptor = outNetwork.Receptors[i];
* receptor.RecurrentCallback = neuronsOnLatLayer[i].Compute;
* receptor.Recurrent = true;
* }
*
* foreach (var neuron in outNetwork.Neurons)
* neuron.LayerId += maxOfAllLayer + 1;
*
*
* receptors.Add(alphaReceptor);
* receptors.AddRange(outNetwork.Receptors);
*
* neurons.AddRange(outNetwork.Neurons);
* synapses.AddRange(outNetwork.Synapses);
*
* return new NeuralNetwork(receptors, neurons, synapses);
*
*/
return(null);
}