public QuickBlackBox(QEAGenome genome)
{
this.genome = genome;
inputs = new double[InputCount];
outputs = new double[OutputCount];
_inputSignalArray = new SignalArray(inputs, 0, inputs.Length);
_outputSignalArray = new SignalArray(outputs, 0, outputs.Length);
}
public QEAGenome createOffspring() // asexual reproduction
{
QEAGenome child = new QEAGenome(this);
for (int i = 0; i < child.weights.Length; i++)
{
for (int j = 0; j < child.weights[i].Length; j++)
{
for (int k = 0; k < child.weights[i][j].Length; k++)
{
if (r.NextDouble() < LEEAParams.MUTATIONRATE)
{
child.weights[i][j][k] += (float)(r.NextDouble() * QuickEvolutionAlgorithm.MUTATIONPOWER * 2 - QuickEvolutionAlgorithm.MUTATIONPOWER);
}
}
}
}
return(child);
}
public QEAGenome createOffspring(QEAGenome parent2) // sexual reproduction
{
QEAGenome child = new QEAGenome(this);
for (int i = 0; i < weights.Length; i++)
{
for (int j = 0; j < weights[i].Length; j++)
{
for (int k = 0; k < weights[i][j].Length; k++)
{
if (r.NextDouble() < 0.5)
{
weights[i][j][k] = parent2.weights[i][j][k];
}
}
}
}
return(child);
}
public BackpropAlgorithm()
{
LEEAParams.DOMAIN.generateSampleList();
evaluator = LEEAParams.DOMAIN.createEvaluator();
// initialize the network, use the QEAGenome class since it's already structured well for this task
int layers = LEEAParams.DOMAIN.getHiddenLayers() + 1; // layers needs to be the number of layers of connections
int[] layerNodes = new int[layers + 1];
layerNodes[0] = LEEAParams.DOMAIN.getInputs() + 1; // +1 for the bias!
for (int i = 0; i < layers - 1; i++)
{
layerNodes[i + 1] = LEEAParams.DOMAIN.getHiddenLayerNeurons(i + 1);
}
layerNodes[layers] = LEEAParams.DOMAIN.getOutputs();
network = new QEAGenome(layers, layerNodes, LEEAParams.BPINITIALWEIGHTMAX);
box = new QuickBlackBox(network);
}
private void createPopulation()
{
genomeList = new List <QEAGenome>();
int layers = LEEAParams.DOMAIN.getHiddenLayers() + 1; // layers needs to be the number of layers of connections
int[] layerNodes = new int[layers + 1];
layerNodes[0] = LEEAParams.DOMAIN.getInputs() + 1; // +1 for the bias!
for (int i = 0; i < layers - 1; i++)
{
layerNodes[i + 1] = LEEAParams.DOMAIN.getHiddenLayerNeurons(i + 1);
}
layerNodes[layers] = LEEAParams.DOMAIN.getOutputs();
Parallel.For(0, LEEAParams.POPSIZE, po, i =>
{
QEAGenome genome = new QEAGenome(layers, layerNodes, 1);
lock (genomeList)
{
genomeList.Add(genome);
}
});
// set up our worker threads
int end;
for (int i = 0; i < workers.Length; i++)
{
end = start + genomeList.Count / LEEAParams.THREADS;
workers[i] = new EvaluationWorker(start, end, genomeList, evaluator);
start = end;
}
// speciate the newly created folks
speciatePopulation();
}
public QEAGenome(QEAGenome copy)
{
r = new FastRandom();
this.weights = copyArray(copy.weights);
}
