public List <Individual> GenerateDescendants(Individual father, Individual mother)
{
Individual mutatedFather = MutateMember(father);
Individual mutatedMother = MutateMember(mother);
if (StaticOperations.ValidateIndividual(mutatedFather) == false)
{
throw new NotSupportedException();
}
if (StaticOperations.ValidateIndividual(mutatedMother) == false)
{
throw new NotSupportedException();
}
List <Individual> crossOverIndividuals = PMX(mutatedMother, mutatedFather);
foreach (Individual individual in crossOverIndividuals)
{
if (StaticOperations.ValidateIndividual(individual) == false)
{
throw new NotSupportedException();
}
}
return(crossOverIndividuals);
}
private List <Individual> CrossOverIndividuals(Individual mutatedMother, Individual mutatedFather)
{
// Descendants are created by randomly switching a randomly generated amount of hours
// Only corresponding hours may be switched and only when both individuals have the same machines running
Individual fatherClone = Individual.CloneRepresentation(mutatedFather);
Individual motherClone = Individual.CloneRepresentation(mutatedMother);
Random random = new Random();
double chance = random.NextDouble();
for (int i = 0; i < fatherClone.Schedule.Count; i++)
{
double value = random.NextDouble();
if (value >= chance)
{
if (IsCrossOverApplicable(fatherClone.Schedule[i], motherClone.Schedule[i]))
{
List <Machine> temp = fatherClone.Schedule[i];
fatherClone.Schedule[i] = motherClone.Schedule[i];
motherClone.Schedule[i] = temp;
}
}
}
ReviseMachineStatus(fatherClone);
ReviseMachineStatus(motherClone);
List <Individual> descendants = new List <Individual>();
if (StaticOperations.ValidateIndividual(motherClone) == false)
{
descendants.Add(mutatedMother);
}
else
{
descendants.Add(motherClone);
}
if (StaticOperations.ValidateIndividual(fatherClone) == false)
{
descendants.Add(mutatedFather);
}
else
{
descendants.Add(fatherClone);
}
return(descendants);
}
public List <Individual> NextGeneration()
{
List <Individual> newGeneration = new List <Individual>();
// Get elite 10% of old population
List <Individual> orderedOldGeneration = oldGeneration.OrderBy(item => item.Fitness).ToList();
orderedOldGeneration.RemoveAll(item => StaticOperations.ValidateIndividual(item) == false);
newGeneration.AddRange(orderedOldGeneration.Take(newGenerationCount / 10));
// Fill up remaining spots with descendants
List <Individual> orderedDescendants = descendants.OrderBy(item => item.Fitness).ToList();
// Remove invalid descendants
orderedDescendants.RemoveAll(item => StaticOperations.ValidateIndividual(item) == false);
newGeneration.AddRange(orderedDescendants.Take(newGenerationCount - newGeneration.Count));
return(newGeneration);
}
public List <Individual> GenerateDescendants(Individual mother, Individual father)
{
Individual mutatedMother = null;
for (int i = 0; i < 5; i++)
{
mutatedMother = MutateIndividual(mother);
if (StaticOperations.ValidateIndividual(mutatedMother))
{
break;
}
}
if (StaticOperations.ValidateIndividual(mutatedMother) == false)
{
mutatedMother = mother;
}
Individual mutatedFather = null;
for (int i = 0; i < 5; i++)
{
mutatedFather = MutateIndividual(father);
if (StaticOperations.ValidateIndividual(mutatedFather))
{
break;
}
}
if (StaticOperations.ValidateIndividual(mutatedFather) == false)
{
mutatedFather = father;
}
List <Individual> descendants = CrossOverIndividuals(mutatedMother, mutatedFather);
if (StaticOperations.ValidateIndividual(descendants[0]) == false)
{
descendants[0] = mother;
}
if (StaticOperations.ValidateIndividual(descendants[1]) == false)
{
descendants[1] = father;
}
return(descendants);
}
public List <Individual> GenerateDescendants(Individual mother, Individual father)
{
Individual motherClone = Individual.CloneIndividual(mother);
Individual fatherClone = Individual.CloneIndividual(father);
double value = random.NextDouble();
Individual mutatedMother = null, mutatedFather = null;
int counter = 0;
for (int i = 0; i < 5; i++)
{
mutatedMother = MutateIndividual(motherClone);
if (StaticOperations.ValidateIndividual(mutatedMother))
{
break;
}
counter++;
}
if (counter == 5)
{
mutatedMother = mother;
}
counter = 0;
for (int i = 0; i < 5; i++)
{
mutatedFather = MutateIndividual(fatherClone);
if (StaticOperations.ValidateIndividual(mutatedFather))
{
break;
}
counter++;
}
if (counter == 5)
{
mutatedFather = father;
}
List <Individual> crossedOvers = CrossOverIndividuals(mutatedMother, mutatedFather);
return(crossedOvers);
}
public List <Individual> GenerateDescendants(Individual mother, Individual father)
{
Individual motherClone = Individual.CloneRepresentation(mother);
Individual fatherClone = Individual.CloneRepresentation(father);
Individual mutatedMother = MutateIndividual(motherClone);
Individual mutatedFather = MutateIndividual(fatherClone);
if (StaticOperations.ValidateIndividual(mutatedMother) == false)
{
mutatedMother = Individual.CloneRepresentation(mother);
}
if (StaticOperations.ValidateIndividual(mutatedFather) == false)
{
mutatedFather = Individual.CloneRepresentation(father);
}
List <Individual> descendants = CrossOverIndividuals(mutatedMother, mutatedFather);
return(descendants);
}
public static void DecoderTesting()
{
List <Point> references = new List <Point>();
Point reference = new Point(3, 21);
references.Add(reference);
reference = new Point(29, 7);
references.Add(reference);
reference = new Point(6, 27);
references.Add(reference);
reference = new Point(8, 18);
references.Add(reference);
reference = new Point(28, 9);
references.Add(reference);
reference = new Point(14, 11);
references.Add(reference);
reference = new Point(9, 13);
references.Add(reference);
reference = new Point(31, 14);
references.Add(reference);
reference = new Point(30, 23);
references.Add(reference);
reference = new Point(23, 25);
references.Add(reference);
reference = new Point(20, 15);
references.Add(reference);
reference = new Point(11, 17);
references.Add(reference);
reference = new Point(27, 12);
references.Add(reference);
reference = new Point(12, 5);
references.Add(reference);
reference = new Point(30, 11);
references.Add(reference);
Individual individual = new Individual(references);
bool status = StaticOperations.ValidateIndividual(individual);
}
private Individual MutateIndividual(Individual individual)
{
Individual individualClone = Individual.CloneRepresentation(individual);
// Every hour needs to be mutated separately
// after mutating one of the hours, values for
// turned on time on machines need to be revised
// Revision is needed only if start/terminate operators are used
Random random = new Random();
for (int i = 0; i < individualClone.Schedule.Count; i++)
{
double mutationChange = random.NextDouble();
if (mutationChange <= 0.25)
{
int mutationEffect = random.Next(0, 3);
// Increase mutation
if (mutationEffect == 0)
{
individualClone.Schedule[i] = IncreaseMutation(individualClone.Schedule[i],
i == 0 ? null : individualClone.Schedule[i - 1], random);
if (StaticOperations.ValidateIndividual(individualClone) == false)
{
individualClone.Schedule[i] = StaticOperations.CloneMachineList(individual.Schedule[i]);
}
}
// Decrease mutation
if (mutationEffect == 1)
{
individualClone.Schedule[i] = DecreaseMutation(individualClone.Schedule[i],
i == 0 ? null : individualClone.Schedule[i - 1], random);
if (StaticOperations.ValidateIndividual(individualClone) == false)
{
individualClone.Schedule[i] = StaticOperations.CloneMachineList(individual.Schedule[i]);
}
}
// Start mutation
if (mutationEffect == 2)
{
individualClone.Schedule[i] = StartMutation(individualClone.Schedule[i],
i == 0 ? null : individualClone.Schedule[i - 1], random);
if (StaticOperations.ValidateIndividual(individualClone) == false)
{
individualClone.Schedule[i] = StaticOperations.CloneMachineList(individual.Schedule[i]);
}
ReviseMachineStatus(individualClone);
}
// Terminate mutation
if (mutationEffect == 3)
{
individualClone.Schedule[i] = TerminateMutation(individualClone.Schedule[i],
i == 0 ? null : individualClone.Schedule[i - 1], random);
if (StaticOperations.ValidateIndividual(individualClone) == false)
{
individualClone.Schedule[i] = StaticOperations.CloneMachineList(individual.Schedule[i]);
}
ReviseMachineStatus(individualClone);
}
}
}
return(individualClone);
}
public void RealiseEvolution()
{
// Initialisation
Initialisation initialisation = new Initialisation(initialPopulationCount, scale);
List <Individual> population = initialisation.GenerateInitialPopulation();
// Validation
foreach (Individual representation in population)
{
if (StaticOperations.ValidateIndividual(representation) == false)
{
throw new NotSupportedException();
}
}
// Evaluation
Evaluation evaluation = new Evaluation(scale);
foreach (Individual representation in population)
{
evaluation.EvaluateIndividual(representation);
}
// Evolution cycles
for (int i = 0; i < evolutionCycles; i++)
{
Console.Write("Epoch #" + i + " ");
// Selection
Selection selection = new Selection(initialPopulationCount, population);
List <Individual> parents = selection.SelectParents(2);
// Genetic operators
List <Individual> descendants = new List <Individual>();
GeneticOperators geneticOperators = new GeneticOperators(scale);
for (int j = 0; j < parents.Count; j = j + 2)
{
descendants.AddRange(geneticOperators.GenerateDescendants(parents[j], parents[j + 1]));
}
// Evaluation
foreach (Individual representation in descendants)
{
evaluation.EvaluateIndividual(representation);
}
// Validation
foreach (Individual representation in population)
{
if (StaticOperations.ValidateIndividual(representation) == false)
{
throw new NotSupportedException();
}
}
// Replacement
Replacement replacement = new Replacement(population, descendants, initialPopulationCount);
population = replacement.NextGeneration();
List <Individual> orderedPopulation = population.OrderBy(item => item.Fitness).ToList();
Console.WriteLine("Best individual has fitness: " + orderedPopulation[0].Fitness);
}
SaveBestIndividualIntoFile(population[0]);
}
public void RealiseEvolution()
{
// Initialise population - validated
Initialisation initialisation = new Initialisation(initialPopulationCount);
List <Individual> population = initialisation.GenerateInitialPopulation();
foreach (Individual individual in population)
{
if (StaticOperations.ValidateIndividual(individual) == false)
{
throw new NotSupportedException();
}
}
// Evaluate synthethic fitness of population
Evaluation evaluation = new Evaluation(firstCriteria, secondCriteria, thirdCriteria);
foreach (Individual individual in population)
{
// Only criteria with true value will be considered
evaluation.EvaluateIndividual(individual);
}
//Evolution cycles
for (int i = 0; i < evolutionCycles; i++)
{
Console.Write("Epoch #" + i);
// Selection - q-tournament
Selection selection = new Selection(population, initialPopulationCount);
List <Individual> parents = selection.SelectParents(2);
// Genetic operators
List <Individual> descendants = new List <Individual>();
GeneticOperators geneticOperators = new GeneticOperators();
for (int j = 0; j < parents.Count; j = j + 2)
{
descendants.AddRange(geneticOperators.GenerateDescendants(parents[j], parents[j + 1]));
}
// Evaluation
foreach (Individual individual in descendants)
{
// Only criteria with true value will be considered
evaluation.EvaluateIndividual(individual);
}
// Replacement
Replacement replacement = new Replacement(population, descendants, initialPopulationCount);
population = replacement.NextGeneration();
// Save best individual
if (firstCriteria)
{
List <Individual> paretSetSurvivors = population.Where(ind => ind.Fitness == 1 && ind.FitnessVector[0] == 1).ToList();
if (paretSetSurvivors.Count == 0)
{
bestIndividuals.Add(population[0]);
}
else
{
paretSetSurvivors.Shuffle();
bestIndividuals.Add(paretSetSurvivors[0]);
}
Console.WriteLine(" Paret set count: " + EvaluateParetSet(population));
}
else
{
List <Individual> paretSetSurvivors = population.Where(ind => ind.Fitness == 1).ToList();
paretSetSurvivors.Shuffle();
bestIndividuals.Add(paretSetSurvivors[0]);
Console.WriteLine(" Paret set count: " + EvaluateParetSet(population));
}
}
SaveBestIndividualsInFile();
}
public void RealiseEvolution()
{
// Initialisation
ReferenceList referenceList = new ReferenceList("tsp.riesenia");
Initialisation initialisation = new Initialisation(initialPopulationCount, referenceList);
List <Individual> population = initialisation.InitialisePopulation();
// Evaluation
Evaluation evaluation = new Evaluation(referenceList);
foreach (Individual individual in population)
{
evaluation.EvaluateIndividual(individual);
}
// Validation
foreach (Individual individual in population)
{
if (StaticOperations.ValidateIndividual(individual) == false)
{
throw new NotSupportedException();
}
}
// Evolution cycles
for (int i = 0; i < evolutionCycles; i++)
{
Console.Write("Epoch #" + i);
// Selection
Selection selection = new Selection(population, population.Count);
List <Individual> parents = selection.SelectParents(2);
// Genetic operators
GeneticOperators geneticOperators = new GeneticOperators();
List <Individual> descendants = new List <Individual>();
for (int j = 0; j < parents.Count; j = j + 2)
{
descendants.AddRange(geneticOperators.GenerateDescendants(parents[j], parents[j + 1]));
}
// Validation
foreach (Individual individual in descendants)
{
if (StaticOperations.ValidateIndividual(individual) == false)
{
throw new NotSupportedException();
}
}
// Evaluation
foreach (Individual individual in descendants)
{
evaluation.EvaluateIndividual(individual);
}
// Replacement
Replacement replacement = new Replacement(population, descendants, population.Count);
population = replacement.NextGeneration();
// Save best individual
List <Individual> orderedPopulation = population.OrderBy(item => item.Fitness).ToList();
bestIndividualPerGeneration.Add(orderedPopulation[0]);
Console.WriteLine(" Minimum fitness: " + orderedPopulation[0].Fitness);
}
SaveBestIndividualsToFile(referenceList);
}
public void RealiseEvolution()
{
// Initialise population
Initialisation initialisation = new Initialisation(initialPopulationCount, goldFieldCount);
List <Individual> population = initialisation.InitialisePopulation();
// Validate population
for (int i = 0; i < population.Count; i++)
{
if (StaticOperations.ValidateIndividual(population[i]) == false)
{
throw new NotSupportedException();
}
}
// Evaluate population
Evaluation evaluation = new Evaluation();
for (int i = 0; i < population.Count; i++)
{
evaluation.EvaluateIndividual(population[i]);
}
// Evolution cycle
for (int i = 0; i < evolutionCycles; i++)
{
Console.Write("# Epoch " + (i + 1));
// Selection
Selection selection = new Selection(population, population.Count);
// Q tournament
List <Individual> parents = selection.SelectParents(4);
// Genetic operators
List <Individual> descendants = new List <Individual>();
GeneticOperators geneticOperators = new GeneticOperators();
for (int j = 0; j < parents.Count; j = j + 2)
{
descendants.AddRange(geneticOperators.GenerateDescendants(parents[j], parents[j + 1]));
}
// Evaluation
for (int j = 0; j < descendants.Count; j++)
{
evaluation.EvaluateIndividual(descendants[j]);
}
// Replacement
Replacement replacement = new Replacement(population, descendants, population.Count);
if (i - bestFitnessEpoch < 100)
{
population = replacement.NextGeneration();
}
else
{
population = replacement.KillBestIndividuals();
bestFitness = double.MaxValue;
}
foreach (Individual individual in population)
{
if (StaticOperations.ValidateIndividual(individual) == false)
{
throw new NotSupportedException();
}
}
// Save best member
List <Individual> orderedPopulation = population.OrderBy(ind => ind.Fitness).ToList();
bestIndividualsPerGeneration.Add(orderedPopulation[0]);
Console.WriteLine(" Minimum fitness: " + orderedPopulation[0].Fitness + ".");
if (orderedPopulation[0].Fitness < bestFitness)
{
bestFitness = orderedPopulation[0].Fitness;
bestFitnessEpoch = i;
}
if (orderedPopulation[0].Fitness == 0)
{
break;
}
}
SaveDataToFile();
}