public Phenotype Crossover(Phenotype other)
{
Random r = new Random();
int splice_point = r.Next(parameters.Input.Size);
Phenotype child = new Phenotype(parameters);
for (int i = 0; i < splice_point; i++)
{
child.Genotype[i] = this.Genotype[i];
}
for (int i = splice_point; i < child.Genotype.Length; i++)
{
if (child.Genotype.Contains(other.Genotype[i]) == false) //ideally - if we can take the second half from the second parent - that's great, let's go for it
{
child.Genotype[i] = other.Genotype[i];
}
else if (child.Genotype.Contains(other.Genotype[i]) == false) //let's see if we can grab the second half from the first parent
{
child.Genotype[i] = this.Genotype[i];
}
else
{ //shove whatever we can
int[] available = Enumerable.Range(1, parameters.Input.Size)
.Where(n => child.Genotype.Contains(n) == false)
.OrderBy(m => r.Next())
.ToArray();
child.Genotype[i] = available[0];
}
}
return(child);
}
override protected Phenotype Selection()
{
int bestRating = BestRating();
int worstRating = WorstRating();
Phenotype chosen = null;
Random r = new Random();
Dictionary <Phenotype, int> selectionPoints = new Dictionary <Phenotype, int>();
int sumOfPoints = 0;
foreach (Phenotype ph in Population)
{
int points = worstRating - parameters.Evaluate(ph.Genotype);
selectionPoints.Add(ph, points);
sumOfPoints += points;
}
int randomized = r.Next(sumOfPoints);
int cumulativeSum = 0;
for (int i = 0; i < Population.Length && chosen == null; i++)
{
cumulativeSum += selectionPoints[Population[i]];
if (randomized < cumulativeSum)
{
chosen = Population[i];
}
}
return(chosen);
}
public override int[] Run()
{
Initialize();
for (int i = 0; i < NOfGenerations; i++)
{
//report results
string GenerationResults = $"{i};{cur_generation.BestRating()};{cur_generation.AverageRating()};{cur_generation.WorstRating()}\n";
sw.Write(GenerationResults);
//next generation
cur_generation = cur_generation.NextGeneration();
if (Evaluate(cur_generation.BestPhenotype().Genotype) < Evaluate(best.Genotype))
{
best = cur_generation.BestPhenotype();
}
}
sw.WriteLine();
for (int i = 0; i < Input.Size; i++)
{
sw.Write($"{best.Genotype[i]}, ");
}
sw.WriteLine(Evaluate(best.Genotype));
sw.Close();
return(best.Genotype);
}
public Generation(GeneticAlgorithm parameters)
{
this.parameters = parameters;
this.Population = new Phenotype[parameters.Population_size];
for (int i = 0; i < parameters.Population_size; i++)
{
Population[i] = new Phenotype(parameters);
}
}
public Generation NextGeneration()
{
Generation nextGeneration;
//brzydkie to to cokolwiek
if (this is GenerationTournament)
{
nextGeneration = new GenerationTournament(this.parameters);
}
else if (this is GenerationRoulette)
{
nextGeneration = new GenerationRoulette(this.parameters);
}
else
{
nextGeneration = null;
throw new Exception("unknown generation type");
}
for (int i = 0; i < Population.Length; i += 2) //using 2 phenotypes from current generation we create 2 for the next one
{
Random r = new Random();
//SELECTION
Phenotype parent_a = Selection();
Phenotype parent_b = Selection();
Phenotype child_a;
Phenotype child_b;
//CROSSOVER
if (r.NextDouble() < parameters.P_crossover)
{
child_a = parent_a.Crossover(parent_b);
child_b = parent_b.Crossover(parent_a);
}
else
{
//TU ZEPSUŁEŚ!
child_a = parent_a.Clone();
child_b = parent_b.Clone();
}
//MUTATION
child_a.Mutate();
child_b.Mutate();
//ADD TO NEW GENERATION
nextGeneration.Population[i] = child_a;
if (i + 1 < Population.Length)
{
nextGeneration.Population[i + 1] = child_b;
}
}
return(nextGeneration);
}
override protected Phenotype Selection()
{
int t_size = parameters.Tournament_size;
Phenotype[] tournament = new Phenotype[t_size];
for (int i = 0; i < t_size; i++)
{
Random r = new Random();
tournament[i] = Population[r.Next(Population.Length)];
}
return(BestPhenotype(tournament));
}
public Phenotype BestPhenotype()
{
Phenotype best = Population[0];
foreach (Phenotype p in Population)
{
if (parameters.Evaluate(p.Genotype) < parameters.Evaluate(best.Genotype))
{
best = p;
}
}
return(best);
}
protected Phenotype BestPhenotype(Phenotype[] population)
{
Phenotype best = population[0];
foreach (Phenotype p in population)
{
if (parameters.Evaluate(p.Genotype) < parameters.Evaluate(best.Genotype))
{
best = p;
}
}
return(best);
}
