public XORPopulation()
{
int popMax = 250;
List <CerebroML.Cerebro> population = new List <CerebroML.Cerebro>();
BrainFactory factory = BrainFactory.Create()
.WithWeightBiasAmplitude(10f)
.WithInput(2)
.WithLayer(2, LayerType.Tanh)
.WithLayer(1, LayerType.Sigmoid);
for (int i = 0; i < popMax; i++)
{
population.Add(factory.Build());
}
this.SetUp(population.ToArray(), 0.05f, 10f);
}
public static void Run()
{
// Generate first population
int popMax = 250;
List <CerebroML.Cerebro> population = new List <CerebroML.Cerebro>();
BrainFactory factory = BrainFactory.Create()
.WithWeightBiasAmplitude(10f)
.WithInput(2)
.WithLayer(2, LayerType.Tanh)
.WithLayer(1, LayerType.Sigmoid);
for (int i = 0; i < popMax; i++)
{
population.Add(factory.Build());
}
// The record
float kingFitness = 0;
CerebroML.Cerebro king = null;
int generation = 0;
// Iterate until the king has a good fitness
while (kingFitness < 0.9) // && generation < GENERATION_LIMT)
{
generation++;
float total = 0;
// Search for the best one
for (int e = 0; e < popMax; e++)
{
float f = CalcFitness(population[e]);
if (f > kingFitness)
{
kingFitness = f;
king = population[e];
}
total += f;
}
// Write to the console the current king and the average fitness
if (generation % 250 == 0)
{
Console.WriteLine($"Gen {generation}");
Console.WriteLine($"Fitness: King- {kingFitness:0.00} AVG- {total / popMax:0.00}");
Console.WriteLine($" - (0, 0) = {king.Run(new float[] { 0.0f, 0.0f })[0]:0.00}");
Console.WriteLine($" - (1, 0) = {king.Run(new float[] { 1.0f, 0.0f })[0]:0.00}");
Console.WriteLine($" - (0, 1) = {king.Run(new float[] { 0.0f, 1.0f })[0]:0.00}");
Console.WriteLine($" - (1, 1) = {king.Run(new float[] { 1.0f, 1.0f })[0]:0.00}");
}
// Make new population
List <CerebroML.Cerebro> newPop = new List <CerebroML.Cerebro>();
for (int e = 0; e < popMax; e++)
{
// Find Two parents to make cross over.
List <CerebroML.Cerebro> parents = new List <CerebroML.Cerebro>();
int control = 0;
Random rnd = new Random();
while (parents.Count < 2 && control < 10000)
{
control++;
int index = rnd.Next(popMax);
float dice = (float)rnd.NextDouble();
if (CalcFitness(population[index]) > dice)
{
parents.Add(population[index]);
}
}
CerebroML.Cerebro offspring = factory.Build();
// Perform cross over and 'inject' the new Genome into the new network
if (parents.Count == 2)
{
Genome e1 = parents[0].GetGenome();
Genome e2 = parents[1].GetGenome();
Genome child = Genome.Crossover(e1, e2);
child.Mutate(0.01f, 10f);
offspring.SetGenome(child);
}
newPop.Add(offspring);
}
population = newPop;
}
// Show the final results
if (king != null)
{
Console.WriteLine("FINAL");
Console.WriteLine($"Fitness: King- {kingFitness:0.00}");
Console.WriteLine($" - (0, 0) = {king.Run(new float[] { 0.0f, 0.0f })[0]:0.00}");
Console.WriteLine($" - (1, 0) = {king.Run(new float[] { 1.0f, 0.0f })[0]:0.00}");
Console.WriteLine($" - (0, 1) = {king.Run(new float[] { 0.0f, 1.0f })[0]:0.00}");
Console.WriteLine($" - (1, 1) = {king.Run(new float[] { 1.0f, 1.0f })[0]:0.00}");
}
}