private AssignmentTwoResultList <GraphGenome> MultiStartLocalSearch(INeighborhood <GraphGenome> neighborhood)
{
//we moeten nog iets met ExperimentAmount doen...
//for(int i = 0; i < ExperimentAmount; ++i)
AssignmentTwoResultList <GraphGenome> results = new AssignmentTwoResultList <GraphGenome>("$MLS_{" + neighborhood.Name + "}$");
for (int j = 0; j < threads.Length; j++)
{
elapsedMilisecondsThreads[j] = new List <AssignmentTwoResults <GraphGenome> >();
int k = j;
int seed = main_random_source.Next();
int amount = OptimaAmount / threads.Length;
Thread T = new Thread(() => searchThread(k, amount, seed, neighborhood));
threads[j] = T;
}
Stopwatch sw = new Stopwatch();
sw.Start();
foreach (Thread T in threads)
{
T.Start();
}
while (this.threadsAreRunning())
{
;
}
sw.Stop();
GraphGenome optimum = null;
for (int j = 0; j < threads.Length; j++)
{
if (optimum == null || optimum.Fitness > solutionsThreads[j].Fitness)
{
optimum = solutionsThreads[j];
}
}
//List<AssignmentTwoResults> results = new List<AssignmentTwoResults>();
foreach (List <AssignmentTwoResults <GraphGenome> > L in elapsedMilisecondsThreads)
{
results.AddRange(L);
}
//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 ... ms", optimum.Fitness); //, results.Sum() / OptimaAmount);
Console.WriteLine("Total time: {0} sec", sw.ElapsedMilliseconds / 1000f);
optimum.ToImage(String.Format("results/MLS[{0}]-{1}.bmp", neighborhood.Name, optimum.Fitness), 3000, 3000);
return(results);
}
public void start(int seed = -1)
{
System.IO.Directory.CreateDirectory("results");
if (seed > 0)
{
main_random_source = new Random(seed);
}
else
{
main_random_source = new Random();
}
INeighborhood <GraphGenome> swap_neighborhood = new SwapNeighborhood <GraphGenome>(main_random_source);
INeighborhood <GraphGenome> FM_neigborhood = new FiducciaMatheysesNeighborhood <GraphGenome>(main_random_source, 7);
// Dan de Fiduccia experimenten:
AssignmentTwoResultList <GraphGenome> ils_FM_results = IteratedLocalSearch(FM_neigborhood);
AssignmentTwoResultList <GraphGenome> mls_FM_results = MultiStartLocalSearch(FM_neigborhood);
AssignmentTwoResultList <GraphGenome> gls_FM_results = GeneticLocalSearch(FM_neigborhood);
// Eerst de "gewone" experimenten:
AssignmentTwoResultList <GraphGenome> mls_results = MultiStartLocalSearch(swap_neighborhood);
AssignmentTwoResultList <GraphGenome> ils_results = IteratedLocalSearch(swap_neighborhood);
AssignmentTwoResultList <GraphGenome> gls_results = GeneticLocalSearch(swap_neighborhood);
string tex_results = "";
tex_results += "\\begin{table}[H]\n\\centering\n\\caption{Comparing 2500 local optima}\n\\label{Comparing 2500 local optima}\n\\begin{tabular}{|l|l|l|l|l|l|l|l|l|}\n";
tex_results += "\\hline\n";
tex_results += "Experiment & Avg. time total & Avg. time 1 optimum & Avg. score & Best score \\\\\n";
tex_results += "\\hline\n";
tex_results += mls_results.OptimaString();
tex_results += ils_results.OptimaString();
tex_results += gls_results.OptimaString();
tex_results += "\\hline\n";
tex_results += mls_FM_results.OptimaString();
tex_results += ils_FM_results.OptimaString();
tex_results += gls_FM_results.OptimaString();
tex_results += "\\hline\n";
tex_results += "\\end{tabular}\n\\end{table}\n";
long max_ticks = gls_results.MinTotalTicks;
tex_results += "\n\n\\begin{table}[H]\n\\centering\n\\caption{Comparing max time of " + max_ticks + " ticks}\n\\label{Comparing max time of " + max_ticks + " ticks}\n\\begin{tabular}{|l|l|l|l|l|l|l|l|l|}\n";
tex_results += "\\hline\n";
tex_results += "Experiment & Avg. \\# optima & Avg. time 1 optimum & Avg. score & Best score \\\\\n";
tex_results += "\\hline\n";
tex_results += mls_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += ils_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += gls_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += "\\hline\n";
tex_results += mls_FM_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += ils_FM_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += gls_FM_results.MaxTicksSubList(max_ticks).MaxTimeString();
tex_results += "\\hline\n";
tex_results += "\\end{tabular}\n\\end{table}\n";
System.IO.StreamWriter file = new System.IO.StreamWriter("results/mls_swap.csv");
file.WriteLine(mls_results.CSV());
file.Close();
file = new System.IO.StreamWriter("results/mls_fm_swap.csv");
file.WriteLine(mls_FM_results.CSV());
file.Close();
file = new System.IO.StreamWriter("results/ils_swap.csv");
file.WriteLine(ils_results.CSV());
file.Close();
file = new System.IO.StreamWriter("results/ils_fm_swap.csv");
file.WriteLine(ils_FM_results.CSV());
file.Close();
file = new System.IO.StreamWriter("results/gls_swap.csv");
file.WriteLine(gls_results.CSV());
file.Close();
file = new System.IO.StreamWriter("results/gls_fm_swap.csv");
file.WriteLine(gls_FM_results.CSV());
file.Close();
double mls_ils_t = t_score(mls_results, ils_results);
double mls_gls_t = t_score(mls_results, gls_results);
double ils_gls_t = t_score(ils_results, gls_results);
double mls_fm_ils_t = t_score(mls_FM_results, ils_FM_results);
double mls_fm_gls_t = t_score(mls_FM_results, gls_FM_results);
double ils_fm_gls_t = t_score(ils_FM_results, gls_FM_results);
double mls_mls_t = t_score(mls_results, mls_FM_results);
double ils_ils_t = t_score(ils_results, ils_FM_results);
double gls_gls_t = t_score(gls_results, gls_FM_results);
tex_results += "\n\n\\begin{table}[H]\n\\centering\n\\caption{Significance of difference between experiments of swap-neighborhood}\n\\label{Significance of difference between experiments of swap-neighborhood}\n\\begin{tabular}{|l|l|l|l|l|l|l|l|l|}\n";
tex_results += "\\hline\n";
tex_results += "Experiment 1 & Experiment 2 & DF & t-score & p-score & significant? \\\\\n";
tex_results += "\\hline\n";
tex_results += "$MLS_{SWAP}$ & $ILS_{SWAP}$ & " + (ExperimentAmount - 1) + " & " + mls_ils_t.ToString() + " & p & x \\\\\n";
tex_results += "$MLS_{SWAP}$ & $GLS_{SWAP}$ & " + (ExperimentAmount - 1) + " & " + mls_gls_t.ToString() + " & p & x \\\\\n";
tex_results += "$ILS_{SWAP}$ & $GLS_{SWAP}$ & " + (ExperimentAmount - 1) + " & " + ils_gls_t.ToString() + " & p & x \\\\\n";
tex_results += "\\hline\n";
tex_results += "\\end{tabular}\n\\end{table}\n";
tex_results += "\n\n\\begin{table}[H]\n\\centering\n\\caption{Significance of difference between experiments of FM-neighborhood}\n\\label{Significance of difference between experiments of FM-neighborhood}\n\\begin{tabular}{|l|l|l|l|l|l|l|l|l|}\n";
tex_results += "\\hline\n";
tex_results += "Experiment 1 & Experiment 2 & DF & t-score & p-score & significant? \\\\\n";
tex_results += "\\hline\n";
tex_results += "$MLS_{FM}$ & $ILS_{FM}$ & " + (ExperimentAmount - 1) + " & " + mls_fm_ils_t.ToString() + " & p & x \\\\\n";
tex_results += "$MLS_{FM}$ & $GLS_{FM}$ & " + (ExperimentAmount - 1) + " & " + mls_fm_gls_t.ToString() + " & p & x \\\\\n";
tex_results += "$ILS_{FM}$ & $GLS_{FM}$ & " + (ExperimentAmount - 1) + " & " + ils_fm_gls_t.ToString() + " & p & x \\\\\n";
tex_results += "\\hline\n";
tex_results += "\\end{tabular}\n\\end{table}\n";
tex_results += "\n\n\\begin{table}[H]\n\\centering\n\\caption{Significance of difference between neighborhoods}\n\\label{Significance of difference between neighborhoods}\n\\begin{tabular}{|l|l|l|l|l|l|l|l|l|}\n";
tex_results += "\\hline\n";
tex_results += "Experiment 1 & Experiment 2 & DF & t-score & p-score & significant? \\\\\n";
tex_results += "\\hline\n";
tex_results += "$MLS_{SWAP}$ & $MLS_{FM}$ & " + (ExperimentAmount - 1) + " & " + mls_mls_t.ToString() + " & p & x \\\\\n";
tex_results += "$ILS_{SWAP}$ & $ILS_{FM}$ & " + (ExperimentAmount - 1) + " & " + ils_ils_t.ToString() + " & p & x \\\\\n";
tex_results += "$GLS_{SWAP}$ & $GLS_{FM}$ & " + (ExperimentAmount - 1) + " & " + gls_gls_t.ToString() + " & p & x \\\\\n";
tex_results += "\\hline\n";
tex_results += "\\end{tabular}\n\\end{table}\n";
string filename = "results/assignment2results.txt";
file = new System.IO.StreamWriter(filename);
file.WriteLine(tex_results);
file.Close();
}
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);
}
