private AssignmentTwoResultList <GraphGenome> GeneticLocalSearch(INeighborhood <GraphGenome> neighborhood)
{
AssignmentTwoResultList <GraphGenome> results = new AssignmentTwoResultList <GraphGenome>("$GLS_{" + neighborhood.Name + "}$");
GraphGenome graph = new GraphGenome(500);
graph.CreateGraph("Graph500.txt");
GraphGenome optimum = null;
for (int i = 0; i < ExperimentAmount; ++i)
{
AssignmentTwoResults <GraphGenome> res = new AssignmentTwoResults <GraphGenome>();
int population_size = 50;
int data_size = 500;
LocalSearch <GraphGenome> local_search =
new LocalSearch <GraphGenome>(data_size,
neighborhood,
new GraphComparer <GraphGenome>(), main_random_source);
LocalSearchProcreator <GraphGenome> lsp = new LocalSearchProcreator <GraphGenome>(
new UniformSymmetricCrossover <GraphGenome>(main_random_source),
local_search,
main_random_source,
res);
GeneticAlgorithm <GraphGenome> ga = new GeneticAlgorithm <GraphGenome>(
data_size,
lsp,
new DefaultSelector <GraphGenome>(
new GraphComparer <GraphGenome>()),
new LocalSearchPopulationGenerator <GraphGenome>(main_random_source, local_search),
new Goal(100, 0),
main_random_source,
"GLS");
InnerResult ir = ga.start(population_size, OptimaAmount / (population_size / 2));
// Eventueel nog een keer uitvoeren om tot 2500 optima te komen.
while (res.Count < OptimaAmount)
{
ir = ga.start(population_size, OptimaAmount / (population_size / 2));
}
// Als we er teveel hebben gekregen door de GLS run meerdere keren uit te voeren,
// pak enkel hoeveel we nodig hebben.
res = res.TakeFirstN(OptimaAmount);
results.Add(res);
if (optimum == null || res.BestResult.Optimum.Fitness < optimum.Fitness)
{
optimum = new GraphGenome(res.BestResult.Optimum.Data, res.BestResult.Optimum.Fitness);
}
}
// We maken iedere keer population_size / 2 optima.
// We willen OptimaAmount optima. Dus we gaan OptimaAmount / (population_size / 2) generaties uitvoeren.
//IteratedLocalSearch();
optimum.ToImage(String.Format("results/GLS[{0}]-{1}.bmp", neighborhood.Name, optimum.Fitness), 3000, 3000);
return(results);
}
private AssignmentTwoResultList <GraphGenome> IteratedLocalSearch(INeighborhood <GraphGenome> neighborhood, bool silent = false)
{
LocalSearch <GraphGenome> local_search =
new LocalSearch <GraphGenome>(500,
neighborhood,
new GraphComparer <GraphGenome>(), main_random_source);
IMutation <GraphGenome> pertubation = new ILSPertubation <GraphGenome>(main_random_source);
GraphGenome graph = new GraphGenome(500);
graph.CreateGraph("Graph500.txt");
GraphGenome optimum = null;
List <long> elapsedMilisecondsList = new List <long>();
AssignmentTwoResultList <GraphGenome> results = new AssignmentTwoResultList <GraphGenome>("$ILS_{" + neighborhood.Name + "}$");
for (int i = 0; i < ExperimentAmount; ++i)
{
AssignmentTwoResults <GraphGenome> res = new AssignmentTwoResults <GraphGenome>();
for (int j = 0; j < OptimaAmount; j++)
{
if (optimum == null)
{
graph = new GraphGenome(500);
graph.Generate(ref main_random_source);
}
else
{
graph = pertubation.Mutate(optimum);
}
Stopwatch sw = Stopwatch.StartNew();
GraphGenome inner_optimum = local_search.Search(graph);
sw.Stop();
elapsedMilisecondsList.Add(sw.ElapsedMilliseconds);
res.Add(inner_optimum, sw.ElapsedTicks);
if (!silent)
{
Console.WriteLine("Found {0} in {1} ticks or {2}ms. ({3}/{4})", inner_optimum.Fitness, sw.ElapsedTicks,
sw.ElapsedMilliseconds, j, OptimaAmount);
}
if (optimum == null || optimum.Fitness >= inner_optimum.Fitness)
{
optimum = inner_optimum;
}
optimum.FunctionEvaluations = 0;
}
results.Add(res);
}
//Console.WriteLine("Best solution found: {0} in an average of {1} ms", optimum.Fitness, cpu_ticks.Average());
Console.WriteLine("Best solution found: {0} in an average of {1} ms", optimum.Fitness, elapsedMilisecondsList.Sum() / OptimaAmount);
//Console.WriteLine("Total time: {0} sec", sw.ElapsedMilliseconds / 1000f);
optimum.ToImage(String.Format("results/ILS[{0}]-{1}.bmp", neighborhood.Name, optimum.Fitness), 3000, 3000);
//Console.ReadLine();
return(results);
}