public override void Select(Population population, SelectionBuffer selection, GeneticAlgorithm.NextStepDelegate callback)
{
int size = 0;
// TODO: Implement settings for how to truncate.
float cutoff = 0;
switch (truncationMode) {
case TruncationMode.AboveMiddle:
cutoff = population.MinFitness + (population.MaxFitness - population.MinFitness)/2;
break;
case TruncationMode.AboveMean:
cutoff = population.MeanFitness;
break;
case TruncationMode.AboveMedian:
cutoff = population.MedianFitness;
break;
}
for (int i = 0; i < population.Size; i++) {
if (population[i].Fitness > cutoff) {
selection[size].CloneFrom(population[i].Genome);
size++;
}
}
selection.Size = size;
callback();
}
public override void Select(Population population, SelectionBuffer selection, GeneticAlgorithm.NextStepDelegate callback)
{
int size = 0;
// TODO: Implement settings for how to truncate.
float cutoff = 0;
switch (truncationMode)
{
case TruncationMode.AboveMiddle:
cutoff = population.MinFitness + (population.MaxFitness - population.MinFitness) / 2;
break;
case TruncationMode.AboveMean:
cutoff = population.MeanFitness;
break;
case TruncationMode.AboveMedian:
cutoff = population.MedianFitness;
break;
}
for (int i = 0; i < population.Size; i++)
{
if (population[i].Fitness > cutoff)
{
selection[size].CloneFrom(population[i].Genome);
size++;
}
}
selection.Size = size;
callback();
}
public override void Mate(SelectionBuffer selected, Population newPopulation, GeneticAlgorithm.NextStepDelegate callback)
{
// The new population is populated with 2 children each iteration.
for (int i = 0; i < newPopulation.Size; i+=2) {
BaseGenome mom = selected[Random.Range(0, selected.Size)]; // TODO: mom and dad can be the same..
BaseGenome dad = selected[Random.Range(0, selected.Size)];
int point = Random.Range(0, mom.Length);
newPopulation[i].Genome.OnePointCrossover(mom, dad, point);
newPopulation[i+1].Genome.OnePointCrossover(mom, dad, point);
}
callback();
}
public override void Mate(SelectionBuffer selected, Population newPopulation, GeneticAlgorithm.NextStepDelegate callback)
{
// The new population is populated with 2 children each iteration.
for (int i = 0; i < newPopulation.Size; i += 2)
{
BaseGenome mom = selected[Random.Range(0, selected.Size)]; // TODO: mom and dad can be the same..
BaseGenome dad = selected[Random.Range(0, selected.Size)];
int point = Random.Range(0, mom.Length);
newPopulation[i].Genome.OnePointCrossover(mom, dad, point);
newPopulation[i + 1].Genome.OnePointCrossover(mom, dad, point);
}
callback();
}
public override void Select(Population population, SelectionBuffer selection, GeneticAlgorithm.NextStepDelegate callback) {
int size = 0;
DebugAux.Assert(population.MaxFitness != 0, "[RouletteWheelSA] Can't have a MaxFitness of zero!");
// The algorithm states that we should pick one at random for consideration with probability 1/N.
int nrConsiderations = Mathf.RoundToInt(population.Size * Settings.SelectionProportion);
for (int i = 0; i < nrConsiderations; i++) {
int index = Random.Range(0, population.Size);
float probability = population[index].Fitness / population.MaxFitness;
if (Random.value <= probability) {
selection[size].CloneFrom(population[index].Genome);
size++;
}
}
selection.Size = size;
callback();
}
public override void Select(Population population, SelectionBuffer selection, GeneticAlgorithm.NextStepDelegate callback)
{
int size = 0;
DebugAux.Assert(population.MaxFitness != 0, "[RouletteWheelSA] Can't have a MaxFitness of zero!");
// The algorithm states that we should pick one at random for consideration with probability 1/N.
int nrConsiderations = Mathf.RoundToInt(population.Size * Settings.SelectionProportion);
for (int i = 0; i < nrConsiderations; i++) {
int index = Random.Range(0, population.Size);
float probability = population[index].Fitness / population.MaxFitness;
if (Random.value <= probability) {
selection[size].CloneFrom(population[index].Genome);
size++;
}
}
selection.Size = size;
callback();
}
// Adaptive GAs
// In CAGA (clustering-based adaptive genetic algorithm)
void Start()
{
settings = GetComponent <GASettings>();
populationSize = settings.PopulationSize;
nrGenerations = settings.NumberOfGenerations;
blocking = settings.RunAsFastAsPossible;
environment.Settings = settings;
selectionStrategy.Settings = settings;
matingStrategy.Settings = settings;
mutationStrategy.Settings = settings;
population = new Population(populationSize);
selectionBuffer = new SelectionBuffer(populationSize);
statistics = GetComponent <Statistics>();
if (statistics == null)
{
statistics = gameObject.AddComponent <Statistics>();
}
statistics.Population = population;
BasePhenome phenomeTemplate = Helper.InstansiateAndGet <BasePhenome>(phenomePrefab);
BaseGenome genomeTemplate = phenomeTemplate.Genome;
for (int i = 0; i < populationSize; i++)
{
BaseGenome genome = genomeTemplate.CreateRandom(); // Should be possible to generate these by hand (seeded)
population[i] = new PhenomeDescription(genome);
}
for (int i = 0; i < populationSize; i++)
{
BaseGenome genome = genomeTemplate.CreateRandom(); // Should be possible to generate these by hand (seeded)
selectionBuffer[i] = genome;
}
nextStepDelegate = NextStep;
currentState = State.FitnessTest;
}
// Should produce two children! public abstract void Mate(SelectionBuffer selected, Population newPopulation, GeneticAlgorithm.NextStepDelegate callback);
// rate only a random sample of the population, as the former process may be very time-consuming. public abstract void Select(Population population, SelectionBuffer selected, GeneticAlgorithm.NextStepDelegate callback);
// Adaptive GAs
// In CAGA (clustering-based adaptive genetic algorithm)
void Start()
{
settings = GetComponent<GASettings>();
populationSize = settings.PopulationSize;
nrGenerations = settings.NumberOfGenerations;
blocking = settings.RunAsFastAsPossible;
environment.Settings = settings;
selectionStrategy.Settings = settings;
matingStrategy.Settings = settings;
mutationStrategy.Settings = settings;
population = new Population(populationSize);
selectionBuffer = new SelectionBuffer(populationSize);
statistics = GetComponent<Statistics>();
if (statistics == null)
statistics = gameObject.AddComponent<Statistics>();
statistics.Population = population;
BasePhenome phenomeTemplate = Helper.InstansiateAndGet<BasePhenome>(phenomePrefab);
BaseGenome genomeTemplate = phenomeTemplate.Genome;
for (int i = 0; i < populationSize; i++) {
BaseGenome genome = genomeTemplate.CreateRandom(); // Should be possible to generate these by hand (seeded)
population[i] = new PhenomeDescription(genome);
}
for (int i = 0; i < populationSize; i++) {
BaseGenome genome = genomeTemplate.CreateRandom(); // Should be possible to generate these by hand (seeded)
selectionBuffer[i] = genome;
}
nextStepDelegate = NextStep;
currentState = State.FitnessTest;
}
// Should produce two children! public abstract void Mate(SelectionBuffer selected, Population newPopulation, GeneticAlgorithm.NextStepDelegate callback);
// rate only a random sample of the population, as the former process may be very time-consuming. public abstract void Select(Population population, SelectionBuffer selected, GeneticAlgorithm.NextStepDelegate callback);
