private double weights(List <newGene> genes1, List <newGene> genes2)
{
//WEIGHTS
Dictionary <int, newGene> i2 = new Dictionary <int, newGene>();
for (int i = 0; i < genes2.Count; i++)
{
newGene gene = genes2[i];
i2[gene.innovation] = gene; //GENE DICTIONARY
}
double sum = 0;
double coincident = 0;
for (int i = 0; i < genes1.Count; i++)
{
newGene gene = genes1[i];
if (i2.Keys.Contains(gene.innovation))
{
newGene gene2 = i2[gene.innovation];
sum = sum + Math.Abs(gene.weight - gene2.weight);
coincident = coincident + 1;
}
}
return(sum / coincident);
}
public Dictionary <string, bool> evaluateNetwork(List <double> inputs)
{
//GET OUTPUT AS BOOLEAN
Inputs = inputs.ToArray(); //SET INPUTS
inputs.Add(1);
if (inputs.Count != InputsCount)
{
Console.WriteLine("Incorrect number of neural network inputs."); return(null);
}
for (int i = 0; i < InputsCount; i++)
{
neurons[i].value = inputs[i];
}
foreach (int neuron in neurons.Keys.ToList()) //LOOP NEVER INPUTS
{
double sum = 0;
for (int i = 0; i < neurons[neuron].incoming.Count; i++)
{
newGene incoming = neurons[neuron].incoming[i];
newNeuron other = neurons[incoming.into];
sum = sum + incoming.weight * other.value; //WEIGHT CALCULATION
}
if (neurons[neuron].incoming.Count > 0)
{
neurons[neuron].value = Sigmoid(sum); //SIGMOID FUNCTION
}
}
Dictionary <string, bool> outputs = new Dictionary <string, bool>();
for (int i = 0; i < OutputsCount; i++)
{
string button = OutputKeys[i]; //GET BUTTON NAMES
if (neurons[MaxNodes + i].value > 0)
{
outputs.Add(button, true); //OUTPUT IS TRUE
}
else
{
outputs.Add(button, false); //OUTPUT IS FALSE
}
Outputs[i] = neurons[MaxNodes + i].value; //SET OUTPUTS
}
//REFRESH IO PANEL
NetMain.IO.setIO(Inputs, Outputs);
NetMain.IO.SecureRefesh();
return(outputs);
}
private bool containsLink(List <newGene> genes, newGene link)
{
//CONTAINS LINK
if (genes != null)
{
for (int i = 0; i < genes.Count; i++) //LOOP GENES
{
newGene gene = genes[i];
if (gene.into == link.into && gene.output == link.output)
{
return(true);
}
}
}
return(false);
}
private double disjoint(List <newGene> genes1, List <newGene> genes2)
{
//DISJOINT
Dictionary <int, bool> i1 = new Dictionary <int, bool>();
//LOOP GENES1
for (int i = 0; i < genes1.Count; i++)
{
newGene gene = genes1[i];
i1[gene.innovation] = true;
}
Dictionary <int, bool> i2 = new Dictionary <int, bool>();
//LOOP GENES2
for (int i = 0; i < genes2.Count; i++)
{
newGene gene = genes2[i];
i2[gene.innovation] = true;
}
//DISJOINT GENES1
int disjointGenes = 0;
for (int i = 0; i < genes1.Count; i++)
{
newGene gene = genes1[i];
if (i2.Keys.Contains(gene.innovation) && !i2[gene.innovation]) //i2.Keys.Contains(gene.innovation) ADD AFTER ERROR (KEY DON'T FIND)
{
disjointGenes = disjointGenes + 1;
}
}
//DISJOINT GENES2
for (int i = 0; i < genes2.Count; i++)
{
newGene gene = genes2[i];
if (i1.Keys.Contains(gene.innovation) && !i1[gene.innovation]) //i1.Keys.Contains(gene.innovation) ADD AFTER ERROR (KEY DON'T FIND)
{
disjointGenes = disjointGenes + 1;
}
}
int n = Max(genes1.Count, genes2.Count);
return((double)disjointGenes / n);
}
private newGenome crossover(newGenome g1, newGenome g2)
{
//MAKE SURE G1 IS THE HIGHER FITNESS GENOME
if (g2.fitness > g1.fitness)
{
newGenome tempg = g1;
g1 = g2;
g2 = tempg;
}
newGenome child = new newGenome();
Dictionary <int, newGene> innovations2 = new Dictionary <int, newGene>();
for (int i = 0; i < g2.genes.Count; i++)
{
newGene gene = g2.genes[i];
innovations2[gene.innovation] = gene;
}
for (int i = 0; i < g1.genes.Count; i++)
{
newGene gene1 = g1.genes[i];
newGene gene2 = innovations2[gene1.innovation];
//COPY GENE1 OR GENE2
if (gene2 != null && Random(1, 2) == 1 && gene2.enabled)
{
child.genes.Add(gene2.Copy());
}
else
{
child.genes.Add(gene1.Copy());
}
}
//GET MAXIMUM
child.maxneuron = Max(g1.maxneuron, g2.maxneuron);
foreach (KeyValuePair <string, double> item in g1.mutationRates)
{
child.mutationRates[item.Key] = item.Value;
}
return(child);
}
private void linkMutate(newGenome genome, bool forceBias)
{
//LINK MUTATE
int neuron1 = randomNeuron(genome.genes, false);
int neuron2 = randomNeuron(genome.genes, true);
newGene newLink = new newGene();
if (neuron1 <= InputsCount && neuron2 <= InputsCount) //BOTH INPUT NODES
{
return;
}
if (neuron2 <= InputsCount) //SWAP OUTPUT AND INPUT
{
int temp = neuron1;
neuron1 = neuron2;
neuron2 = temp;
}
newLink.into = neuron1;
newLink.output = neuron2;
if (forceBias)
{
newLink.into = InputsCount;
}
if (containsLink(genome.genes, newLink))
{
return;
}
newLink.innovation = newInnovation(Pool);
newLink.weight = Random() * 4 - 2;
//ADD TO GENES LIST
genome.genes.Add(newLink);
}
private void nodeMutate(newGenome genome)
{
//NODE MUTATE
if (genome.genes.Count == 0)
{
return;
}
genome.maxneuron = genome.maxneuron + 1;
//GET RANDOM GENE
newGene gene = genome.genes[Random(0, genome.genes.Count - 1)];
if (!gene.enabled) //IF FALSE RETURN
{
return;
}
gene.enabled = false;
//ADD GENE1
newGene gene1 = gene.Copy();
gene1.output = genome.maxneuron;
gene1.weight = 1.0;
gene1.innovation = newInnovation(Pool);
gene1.enabled = true;
genome.genes.Add(gene1);
//ADD GENE2
newGene gene2 = gene.Copy();
gene2.into = genome.maxneuron;
gene2.innovation = newInnovation(Pool);
gene2.enabled = true;
genome.genes.Add(gene2);
}
private void enableDisableMutate(newGenome genome, bool enable)
{
//ENABLE DISABLE MUTATE
List <newGene> candidates = new List <newGene>();
//GET ENABLES GENES
foreach (newGene item in genome.genes)
{
if (item.enabled == !enable)
{
candidates.Add(item);
}
}
if (candidates.Count == 0)
{
return;
}
//GET RANDOM GENE FROM CANIDATES
newGene gene = candidates[Random(0, candidates.Count - 1)];
gene.enabled = !gene.enabled;
}
public generateNetwork(newGenome genome)
{
//INPUTS
for (int i = 0; i < InputsCount; i++)
{
neurons[i] = new newNeuron();
}
//OUTPUTS
for (int i = 0; i < OutputsCount; i++)
{
neurons[MaxNodes + i] = new newNeuron();
}
genome.genes.Sort(); //table.sort(genome.genes, function(a, b) return (a.output < b.output) end)
for (int i = 0; i < genome.genes.Count; i++)
{
newGene gene = genome.genes[i];
if (gene.enabled)
{
if (!neurons.Keys.Contains(gene.output))
{
neurons[gene.output] = new newNeuron();
}
newNeuron neuron = neurons[gene.output];
neuron.incoming.Add(gene);
if (!neurons.Keys.Contains(gene.into))
{
neurons[gene.into] = new newNeuron();
}
}
genome.network = this;
}
}
public void Load()
{
//LOAD
string[] path = Mod_File.FileOpenDialog(false, FILTER.TXT);
//ABBRUCH
if (path == null)
{
return;
}
//START LOADING
UniLoad.loadingStart();
//READ TXT FILE
string[] file = Mod_TXT.readTXT(path[0]);
int x = 0;
//GET DURATION
Duration = TimeSpan.FromMilliseconds(Mod_Convert.StringToDouble(file[x++]));
//GET OUTPUT KEYS
OutputKeys = file[x++].Split(' ');
//INITIALIZE POOL
Initialize(OutputKeys, true);
Pool.generation = Mod_Convert.StringToInteger(file[x++]);
Pool.maxFitness = Mod_Convert.StringToDouble(file[x++]);
int numSpecies = Mod_Convert.StringToInteger(file[x++]);
for (int j = 0; j < numSpecies; j++) //SPECIES
{
newSpecies species = new newSpecies();
Pool.species.Add(species);
species.topFitness = Mod_Convert.StringToDouble(file[x++]);
species.staleness = Mod_Convert.StringToInteger(file[x++]);
int numGenomes = Mod_Convert.StringToInteger(file[x++]);
for (int i = 0; i < numGenomes; i++) //GENOME
{
newGenome genome = new newGenome();
species.genomes.Add(genome);
genome.fitness = Mod_Convert.StringToDouble(file[x++]);
genome.maxneuron = Mod_Convert.StringToInteger(file[x++]);
string line = file[x++];
while (line != "done")
{
genome.mutationRates[line] = Mod_Convert.StringToDouble(file[x++]);
line = file[x++];
}
int numGenes = Mod_Convert.StringToInteger(file[x++]);
for (int k = 0; k < numGenes; k++) //GENE
{
newGene gene = new newGene();
genome.genes.Add(gene);
string[] split = file[x++].Split(' ');
gene.into = Mod_Convert.StringToInteger(split[0]);
gene.output = Mod_Convert.StringToInteger(split[1]);
gene.weight = Mod_Convert.StringToDouble(split[2]);
gene.innovation = Mod_Convert.StringToInteger(split[3]);
gene.enabled = Mod_Convert.ObjectToBool(split[4]);
}
}
}
//FITNESS ALREADY MEASURED
while (fitnessAlreadyMeasured())
{
nextGenome();
}
initializeRun();
//UPDATE LEARN PANEL
NetMain.RoundFinished(0);
//END LOADING
UniLoad.loadingEnd();
}