public CandidateSolutionItem(int capacity, ItemList list, IRandomizer randomizer)
{
_randomizer = randomizer;
CandidateItem = list;
Capacity = capacity;
Genes = new List <bool>(list.Count);
for (int i = 0; i < list.Count; i++)
{
Genes.Add(randomizer.GetDoubleFromZeroToOne() >= 0.5);
}
Repair();
CalcFitness();
}
public T Run(TInput data)
{
CurrentGeneration = new List <ICandidateSolution <T> >(PopulationSize);
for (int i = 0; i < PopulationSize; i++)
{
CurrentGeneration.Add(_generator.GenerateCandidate(data));
}
ICandidateSolution <T> bestSolution = null;
int bestFitness = int.MinValue;
var bestGen = 0;
var gen = 1;
while (true)
{
var bestGenFitness = 0;
var totalFitness = 0;
ICandidateSolution <T> bestSolutionThisGen = null;
foreach (var candidateSolution in CurrentGeneration)
{
candidateSolution.Repair();
totalFitness += candidateSolution.Fitness;
if (candidateSolution.Fitness > bestGenFitness)
{
bestGenFitness = candidateSolution.Fitness;
bestSolutionThisGen = candidateSolution;
}
}
if (totalFitness > bestFitness)
{
bestFitness = totalFitness;
bestSolution = bestSolutionThisGen;
bestGen = gen;
}
else
{
if ((gen - bestGen) > MaxGenerationsWithNoChanges)
{
break;
}
}
var nextGen = new List <ICandidateSolution <T> >();
while (nextGen.Count < PopulationSize)
{
var parent1 = RouletteWheelSelectCandidate(totalFitness).CloneObject();
var parent2 = RouletteWheelSelectCandidate(totalFitness).CloneObject();
if (_randomizer.GetDoubleFromZeroToOne() < CrossOverRate)
{
var(child1, child2) = _generator.CrossOver(parent1, parent2, data);
parent1 = child1;
parent2 = child2;
}
if (_randomizer.GetDoubleFromZeroToOne() < MutationRate)
{
parent1 = _generator.Mutate(parent1, data, MutationRate);
}
if (_randomizer.GetDoubleFromZeroToOne() < MutationRate)
{
parent2 = _generator.Mutate(parent2, data, MutationRate);
}
nextGen.Add(parent1);
nextGen.Add(parent2);
}
gen++;
}
return(bestSolution?.CandidateItem);
}