public GeneticNeuronGenerator(
INeuronActivator neuronActivator,
int populationSize,
int[] topology,
params IGenerationUpdater[] generationUpdaters)
{
_neuralNetworkTopology = topology;
_genotypeGenesAmount = NeuralNetwork.CalculateWeightCount(topology);
_currentPopulation = new GeneticNeuron[populationSize];
EvolutionManager = new EvolutionManager(_genotypeGenesAmount, populationSize, generationUpdaters);
_neuronActivator = neuronActivator;
}
public static NeuralNetwork CreateNeuralNetworkFromFile(string path, INeuronActivator neuronActivator)
{
XDocument doc = XDocument.Load(path);
if (doc.Root == null)
{
throw new Exception("Document is empty");
}
XElement root = doc.Root;
var biasList = new List <string>(8);
var weightList = new List <string>(8);
var inputCountList = new List <string>(8);
var outputCountList = new List <string>(8);
foreach (XElement layerElement in root.Elements(LayerId))
{
biasList.Add(layerElement.Attribute(BiasId)?.Value);
inputCountList.Add(layerElement.Attribute(InputCountId)?.Value);
outputCountList.Add(layerElement.Attribute(OutputCountId)?.Value);
foreach (XElement element in layerElement.Elements())
{
weightList.Add(element.Value);
}
}
int[] topology = new int[inputCountList.Count + 1];
for (int i = 0; i < inputCountList.Count; i++)
{
topology[i] = int.Parse(inputCountList[i]);
}
topology[topology.Length - 1] = int.Parse(outputCountList[outputCountList.Count - 1]);
float[] weights = weightList.Select(float.Parse).ToArray();
var network = new NeuralNetwork(neuronActivator, topology);
network.FillLayers(weights);
for (int i = 0; i < biasList.Count; i++)
{
string biasString = biasList[i];
if (!string.IsNullOrEmpty(biasString) && float.TryParse(biasString, out float bias))
{
network.Layers[i].Bias = bias;
}
}
return(network);
}
public NeuralLayer(int inputCount, int outputCount, INeuronActivator neuronActivator = null)
{
_neuronActivator = neuronActivator;
InputCount = inputCount;
OutputCount = outputCount;
Weights = new float[InputCount + 1][];
for (int i = 0; i < Weights.Length; i++)
{
Weights[i] = new float[OutputCount];
}
_biasedInputs = new float[InputCount + 1];
}
public NeuralNetwork(INeuronActivator neuronActivator, params int[] topology)
{
#if DEBUG
if (topology.Length <= 1)
{
throw new Exception("Network should have at least 1 layer");
}
#endif
TotalWeightCount = CalculateWeightCount(topology);
Layers = new NeuralLayer[topology.Length - 1];
_layersOutput = new float[Layers.Length][];
for (int i = 0; i < Layers.Length; i++)
{
int outputCount = topology[i + 1];
Layers[i] = new NeuralLayer(topology[i], outputCount, neuronActivator);
_layersOutput[i] = new float[outputCount];
}
}
