public IList <IChromosome> Cross(IPopulation population, ICrossover crossover, float crossoverProbability, IList <IChromosome> parents)
{
var minSize = population.MinSize;
var offspring = new List <IChromosome>(minSize);
for (int i = 0; i < minSize; i += crossover.ParentsNumber)
{
var selectedParents = parents.Skip(i).Take(crossover.ParentsNumber).ToList();
// If match the probability cross is made, otherwise the offspring is an exact copy of the parents.
// Checks if the number of selected parents is equal which the crossover expect, because the in the end of the list we can
// have some rest chromosomes.
if (selectedParents.Count == crossover.ParentsNumber && RandomizationProvider.Current.GetDouble() <= crossoverProbability)
{
offspring.AddRange(crossover.Cross(selectedParents));
}
else
{
var left = crossover.ParentsNumber - selectedParents.Count;
for (int j = 0; j < left; j++)
{
offspring.Add(parents[RandomizationProvider.Current.GetInt(0, parents.Count)].Clone());
}
}
}
return(offspring);
}
public IList <IChromosome> Cross(IPopulation population, ICrossover crossover, float crossoverProbability, IList <IChromosome> parents)
{
var offspring = new List <IChromosome>();
var clonedParents = parents.Select(x => x.Clone()).ToList();
for (int i = 0; i < population.MaxSize; i += crossover.ParentsNumber)
{
var selectedParents = clonedParents.Skip(i).Take(crossover.ParentsNumber).ToList();
// If match the probability cross is made, otherwise the offspring is an exact copy of the parents.
// Checks if the number of selected parents is equal which the crossover expect, because in the end of the list we can
// have some rest chromosomes.
if (selectedParents.Count == crossover.ParentsNumber)
{
bool shouldCrossover = RandomizationProvider.Current.GetDouble() <= crossoverProbability;
offspring.AddRange(shouldCrossover ?
crossover.Cross(selectedParents) : selectedParents);
}
else
{
offspring.AddRange(selectedParents);
int leftToAdd = crossover.ParentsNumber - selectedParents.Count;
for (int j = 0; j < leftToAdd; j++)
{
offspring.Add(clonedParents[RandomizationProvider.Current.GetInt(0, clonedParents.Count)].Clone());
}
}
}
return(offspring);
}
/// <summary>
/// Enumerates through the offspring created from the specie.
/// </summary>
/// <param name="population"></param>
/// <param name="specie"></param>
/// <param name="crossover"></param>
/// <param name="childrenCount"></param>
/// <param name="allParents"></param>
/// <param name="parentsLookup"></param>
/// <returns></returns>
protected virtual IEnumerable <IChromosome> PerformCrossSpecie(
ISpeciedPopulation population,
Specie specie,
ICrossover crossover,
int childrenCount,
IList <IChromosome> allParents,
ILookup <Specie, IChromosome> parentsLookup)
{
if (parentsLookup[specie].Count() < crossover.ParentsNumber)
{
throw new InvalidOperationException("Not enough specie parents available for this crossover.");
}
if (childrenCount <= 0)
{
yield break;
}
var curParents = new List <IChromosome>(crossover.ParentsNumber);
int pInd = 0;
for (int i = 0; i < childrenCount; i += crossover.ChildrenNumber)
{
IList <IChromosome> kids;
// Interspecie crossover
if (RandomizationProvider.Current.GetDouble() <= InterspecieMatingRate)
{
kids = PerformInterspecieCross(crossover, allParents);
}
// Crossover within specie
else
{
List <IChromosome> specieParents = parentsLookup[specie].ToList();
// Choose parents, place into curParents list
curParents.Clear();
while (curParents.Count < crossover.ParentsNumber)
{
curParents.Add(specieParents[pInd++]);
// Wrap around and re-shuffle
if (pInd >= specieParents.Count)
{
pInd = 0;
specieParents = specieParents.Shuffle(RandomizationProvider.Current).ToList();
}
}
kids = crossover.Cross(curParents);
}
// Return children
for (int k = 0; k < kids.Count; k++)
{
if (i + k < childrenCount)
{
yield return(kids[k]);
}
}
}
}
/// <summary>
/// Performs a crossover by choosing parents uniformly from the list of parents.
/// </summary>
protected IList <IChromosome> PerformInterspecieCross(ICrossover crossover, IList <IChromosome> allParents)
{
var ints = RandomizationProvider.Current.GetUniqueInts(crossover.ParentsNumber, 0, allParents.Count);
var curParents = new IChromosome[crossover.ParentsNumber];
for (int i = 0; i < crossover.ParentsNumber; i++)
{
curParents[i] = allParents[ints[i]];
}
return(crossover.Cross(curParents));
}
public Individual[] ProcessEpoche()
{
Individual[] newPopulation = new Individual[population.Length];
for (int i = 0; i < population.Length; i++)
{
Individual[] parents = FindParents();
Individual child = crossoverOperator.Cross(parents, cities);
Individual mutatedChild = methodOfMutation.Mutate(cities, child);
newPopulation[i] = mutatedChild;
}
population = newPopulation;
return(population);
}
/// <summary>
/// form parent matches and either executes the corresponding crossover with given probability to produce new children, or returns null children
/// </summary>
/// <param name="population">the population from which parents are selected</param>
/// <param name="crossover">The crossover class.</param>
/// <param name="crossoverProbability">The crossover probability.</param>
/// <param name="parents">The parents.</param>
/// <param name="firstParentIndex">the index of the first parent selected for a crossover</param>
/// <returns>children for the current crossover if it was performed, null otherwise</returns>
protected IList <IChromosome> SelectParentsAndCross(IPopulation population, ICrossover crossover,
float crossoverProbability, IList <IChromosome> parents, int firstParentIndex)
{
var selectedParents = parents.Skip(firstParentIndex).Take(crossover.ParentsNumber).ToList();
// If match the probability cross is made, otherwise the offspring is an exact copy of the parents.
// Checks if the number of selected parents is equal which the crossover expect, because the in the end of the list we can
// have some rest chromosomes.
if (selectedParents.Count == crossover.ParentsNumber && RandomizationProvider.Current.GetDouble() <= crossoverProbability)
{
return(crossover.Cross(selectedParents));
}
return(null);
}
public Tour(Generation generation, ITwoElementsStrategy strategy, ICrossover crossover)
{
task = generation[0].Task;
int index1, index2;
strategy.Run(generation, out index1, out index2);
Tour parent1 = generation[index1];
Tour parent2 = generation[index2];
parent1.CorrectRotation();
parent2.CorrectRotation();
vertices = crossover.Cross(parent1, parent2);
UpdateLength();
}
public void Run_WithChromsomeMatchingTerminationCriteria_CallsRightMethods()
{
var deploymentModel = CreateDeploymentModel();
A.CallTo(() => _terminationCondition.HasReached(null, null)).WithAnyArguments().Returns(true);
ConstructGeneticAlgorithm();
var bestDeploymentModel = _testGeneticAlgorithm.Run(deploymentModel);
A.CallTo(() => _deploymentChromosomeFactory.Create(deploymentModel)).MustHaveHappened(Repeated.Exactly.Once);
A.CallTo(() => _initialPopulationCreator.CreateInitialPopulation(null, 0, 0)).WithAnyArguments().MustHaveHappened(Repeated.Exactly.Once);
A.CallTo(() => _mutationOperator.Mutate(null)).WithAnyArguments().MustNotHaveHappened();
A.CallTo(() => _crossoverOperator.Cross(null)).WithAnyArguments().MustNotHaveHappened();
A.CallTo(() => _selectionStrategy.SelectChromosomes(0, null)).WithAnyArguments().MustNotHaveHappened();
A.CallTo(() => _reinsertionStrategy.Reinsert(null, null)).WithAnyArguments().MustNotHaveHappened();
Assert.AreEqual(0, _currentState.Count);
}
/// <summary>
/// Performs crossover of the two schedule chromosomes.
/// </summary>
/// <param name="parents"></param>
/// <returns></returns>
/// <exception cref="ArgumentException"></exception>
public IList <IChromosome> Cross(IList <IChromosome> parents)
{
var scheduleParent1 = parents[0];
var scheduleParent2 = parents[1];
if (scheduleParent1.Length != scheduleParent2.Length)
{
throw new ArgumentException("Different length of parents, cannot be crossed.");
}
var scheduleChild1 = (ScheduleChromosome)scheduleParent1.Clone();
var scheduleChild2 = (ScheduleChromosome)scheduleParent2.Clone();
for (int i = 0; i < scheduleParent1.Length; i++)
{
var machine1 = (MachineChromosome)scheduleParent1.GetGene(i).Value;
var machine2 = (MachineChromosome)scheduleParent2.GetGene(i).Value;
if (machine1.Length < 3)
{
continue;
}
var result = permutationCrossover.Cross(new List <IChromosome>()
{
machine1, machine2
});
var child1 = (MachineChromosome)result[0];
var child2 = (MachineChromosome)result[1];
// replace genes (automatically resets fitness)
scheduleChild1.ReplaceGene(i, new Gene(child1));
scheduleChild1.ScheduleLength = 0;
scheduleChild2.ReplaceGene(i, new Gene(child2));
scheduleChild2.ScheduleLength = 0;
}
return(new List <IChromosome>()
{
scheduleChild1, scheduleChild2
});
}
private IList <ChromoSome> _Cross(IList <ChromoSome> parents)
{
var count = _crossover.ParentCount;
var subParents = new List <ChromoSome>();
var children = new List <ChromoSome>(_population.MinPopulationCount);
foreach (var parent in parents)
{
subParents.Add(parent);
if (subParents.Count == count)
{
if (Util.NextDouble() < _crossoverProbability)
{
children.AddRange(_crossover.Cross(subParents));
}
subParents.Clear();
}
}
return(children);
}
public Individual <T> Select(List <Individual <T> > population)
{
var randomized = new int[_k];
for (var i = 0; i < _k; i++)
{
var index = Random.Next(population.Count);
randomized[i] = index;
}
randomized = randomized.OrderBy(i => population[i].Fitness).ToArray();
var best = population[randomized[0]];
var secondBest = population[randomized[1]];
var worst = population[randomized[2]];
var child = _crossover.Cross(best, secondBest);
// A possible scientific breakthrough. The error was less than 1E-41 on 80 dimensions for F7.
// _mutation.Mutate(child, _lowerBound, _upperBound, _mutationProbability, _mean, _deviation);
_mutation.Mutate(child, _mean, _deviation, _lowerBound, _upperBound, _mutationProbability);
child.Fitness = EvaluateIndividual(_function, child);
if (child.Fitness > worst.Fitness)
{
population[randomized[2]] = child;
}
// This yields much better results
// my mind is telling me no
// but my body.. my body is telling me yes
if (child.Fitness > best.Fitness)
{
population[randomized[0]] = child;
}
return(child);
}
/// <summary>
/// Crosses the specified parents.
/// </summary>
/// <param name="crossover">The crossover class.</param>
/// <param name="crossoverProbability">The crossover probability.</param>
/// <param name="parents">The parents.</param>
/// <returns>The result chromosomes.</returns>
public IList <IChromosome> Cross(IPopulation population, ICrossover crossover, float crossoverProbability, IList <IChromosome> parents)
{
var offspring = new ConcurrentBag <IChromosome>();
Parallel.ForEach(Enumerable.Range(0, population.MinSize / crossover.ParentsNumber).Select(i => i * crossover.ParentsNumber), i =>
{
var selectedParents = parents.Skip(i).Take(crossover.ParentsNumber).ToList();
// If match the probability cross is made, otherwise the offspring is an exact copy of the parents.
// Checks if the number of selected parents is equal which the crossover expect, because the in the end of the list we can
// have some rest chromosomes.
if (selectedParents.Count == crossover.ParentsNumber && RandomizationProvider.Current.GetDouble() <= crossoverProbability)
{
var children = crossover.Cross(selectedParents);
foreach (var item in children)
{
offspring.Add(item);
}
}
});
return(offspring.ToList());
}
