//The Step function assumes that the fitness values of all the individuals in the population have been calculated.
public override void Step()
{
//criar a nova populacao
List <Individual> new_pop = new List <Individual>();
updateReport(); //called to get some stats
// fills the rest with mutations of the best!
if (elitist)
{
for (int i = 0; i < populationSize - (elistismoValor * populationSize); i++)
{
GeneticIndividual best = (GeneticIndividual) new TournmentSelect().selectIndividuals(population, tournamentSize).Clone();
GeneticIndividual best2 = (GeneticIndividual) new TournmentSelect().selectIndividuals(population, tournamentSize).Clone();
best.Crossover(best2, crossoverProbability);
best.Mutate(mutationProbability);
new_pop.Add(best.Clone());
Debug.Log(new_pop.Count);
}
float max2 = float.MaxValue;
Individual elite = null;
for (int i = 0; i < (elistismoValor * populationSize); i++)
{
float max = float.MinValue;
for (int j = 0; j < populationSize; j++)
{
if (population[j].Fitness > max && population[j].Fitness < max2)
{
max = population[j].Fitness;
elite = (GeneticIndividual)population[j];
}
}
max2 = max;
new_pop.Add(elite.Clone());
}
}
else
{
for (int i = 0; i < populationSize; i++)
{
GeneticIndividual best = (GeneticIndividual) new TournmentSelect().selectIndividuals(population, tournamentSize).Clone();
GeneticIndividual best2 = (GeneticIndividual) new TournmentSelect().selectIndividuals(population, tournamentSize).Clone();
best.Crossover(best2, crossoverProbability);
best.Mutate(mutationProbability);
new_pop.Add(best.Clone());
Debug.Log(new_pop.Count);
}
}
population = new_pop;
generation++;
}
public override void Step()
{
List <Individual> new_pop = new List <Individual>();
updateReport(); //called to get some stats
if (elitist)
{
GeneticIndividual melhor = (GeneticIndividual)overallBest.Clone();
new_pop.Add(melhor); //coloca o melhor de sempre na populaçao
}
while (new_pop.Count < populationSize)
{
GeneticIndividual parent1 = (GeneticIndividual)selection.torneio(population, tournamentSize); //torneio
GeneticIndividual parent2;
do
{
parent2 = (GeneticIndividual)selection.torneio(population, tournamentSize);
} while (parent1 == parent2);//para que o pai 1 seja diferente do 2
GeneticIndividual tmp1 = (GeneticIndividual)parent1.Clone();
GeneticIndividual tmp2 = (GeneticIndividual)parent2.Clone();
tmp1.Crossover(tmp2, crossoverProbability); // crossover em temporarios
tmp2.Crossover(tmp1, crossoverProbability);
GeneticIndividual child1 = tmp1;// apenas para clareza
GeneticIndividual child2 = tmp2;
child1.Mutate(mutationProbability);//mutaçoes
child2.Mutate(mutationProbability);
new_pop.Add(child1); //adiciona à população os filhos
new_pop.Add(child2);
}
population = new_pop;
/*
* float maximum = population[0].Fitness;
* for(int i = 0; i < population.Count - 1; i++)
* {
* if (population[i].Fitness > maximum)
* {
* maximum = population[i].Fitness;
* }
* }
*
* Debug.Log("Max: " + maximum);
*/
generation++;//passa para a próxima geração
}
//The Step function assumes that the fitness values of all the individuals in the population have been calculated.
public override void Step()
{
System.Random random = new System.Random();
//You should implement the code runs in each generation here
List <Individual> new_pop = new List <Individual>();
updateReport();
population.Sort((x, y) => x.Fitness.CompareTo(y.Fitness));
population.Reverse();
if (elitist == true)
{
for (int i = 0; i < elitistSize; i++)
{
new_pop.Add(population[i].Clone());
}
}
if (tournamentSize > 0)
{
List <Individual> winners = fitnessSelect.selectIndividuals(population, tournamentSize, tournamentProbability);
for (int i = new_pop.Count; i < population.Count; i++)
{
Individual parent1 = winners[Random.Range(0, winners.Count)];
winners.Remove(parent1);
Individual parent2 = winners[Random.Range(0, winners.Count)].Clone();
parent2.Crossover(parent1, crossoverProbability);
parent2.Mutate(mutationProbability);
winners.Add(parent1);
new_pop.Add(parent2);
}
}
else
{
for (int i = new_pop.Count; i < populationSize; i++)
{
GeneticIndividual tmp = (GeneticIndividual)overallBest.Clone();
tmp.Mutate(mutationProbability);
new_pop.Add(tmp);
}
}
population = new_pop;
generation++;
}
public override void InitPopulation()
{
maxNumberOfEvaluations = Mathf.Min(maxNumberOfEvaluations, populationSize);
populationRed = new List <Individual>();
populationBlue = new List <Individual>();
while (populationRed.Count < populationSize)
{
GeneticIndividual new_ind_red = new GeneticIndividual(NNTopology, maxNumberOfEvaluations, mutationMethod)
{
Mean = Mean,
Stdev = Stdev
};
GeneticIndividual new_ind_blue = new GeneticIndividual(NNTopology, maxNumberOfEvaluations, mutationMethod)
{
Mean = Mean,
Stdev = Stdev
};
if (seedPopulationFromFile)
{
NeuralNetwork nnRed = getRedIndividualFromFile();
NeuralNetwork nnBlue = getBlueIndividualFromFile();
new_ind_red.Initialize(nnRed);
new_ind_blue.Initialize(nnBlue);
//only the first individual is an exact copy. the other are going to suffer mutations
if (populationRed.Count != 0 && populationBlue.Count != 0)
{
new_ind_red.Mutate(mutationProbabilityRedPopulation);
new_ind_blue.Mutate(mutationProbabilityBluePopulation);
}
}
else
{
new_ind_red.Initialize();
new_ind_blue.Initialize();
}
populationRed.Add(new_ind_red);
populationBlue.Add(new_ind_blue);
}
switch (selectionMethod)
{
case SelectionType.Tournament:
selection = new TournamentSelection(tournamentSize);
break;
}
}
//The Step function assumes that the fitness values of all the individuals in the population have been calculated.
public override void Step()
{
List <Individual> new_pop = new List <Individual> ();
List <Individual> elitist_list = new List <Individual> ();
if (elitist)
{
population.Sort((x, y) => x.Fitness.CompareTo(y.Fitness));
Debug.Log("P: ");
for (int i = 0; i < populationSize; i++)
{
Debug.Log(population [i].Fitness);
}
for (int i = 0; i < n_elitists; i++)
{
int pop = populationSize - 1;
elitist_list.Add(population [pop - i].Clone());
}
}
updateReport(); //called to get some stats
// fills the rest with mutations of the best !
for (int i = 0; i < populationSize; i++)
{
GeneticIndividual pai = (GeneticIndividual)Tournament(population, tournamentSize).Clone();
GeneticIndividual pai2 = (GeneticIndividual)Tournament(population, tournamentSize).Clone();
pai.Crossover(pai2.Clone(), crossoverProbability, numerocortes);
pai.Mutate(mutationProbability);
new_pop.Add(pai.Clone());
}
population = new_pop;
if (elitist)
{
for (int i = 0; i < n_elitists; i++)
{
Individual pick_to_delete = new_pop [Random.Range(0, populationSize - 1)];
new_pop.Remove(pick_to_delete);
new_pop.Add(elitist_list [i].Clone());
}
}
generation++;
}