public static async Task <IConfiguration> TestConfigurations(IConfiguration first_configuration, IConfiguration second_configuration, int number_of_measurements = 30, int evaluation_duration = 400, int seed_limit = 100)
{
if (number_of_measurements > seed_limit)
{
throw new Exception("Error: El número de mediciones no puede superar a la semilla límite.");
}
var random = new Random();
var seeds = Enumerable.Range(1, seed_limit).OrderBy(num => random.Next()).Take(number_of_measurements).ToList();
var u_values = new List <U_Value>();
Table = new DynamicTable(Console.CursorTop, "U de Mann Whitney", new string[] { "Medición", first_configuration.ToString(), second_configuration.ToString() }, seeds.Select((seed, index) => new string[] { (index + 1).ToString(), "-", "-" }).ToArray());
Table.Display("Nivel de significancia α: 5%");
var get_measurements_tasks = new List <Task <List <U_Value> > > {
Task.Run(() => GetMeasurements(first_configuration, 1, evaluation_duration, seeds)),
Task.Run(() => GetMeasurements(second_configuration, 2, evaluation_duration, seeds))
};
await Task.WhenAll(get_measurements_tasks);
get_measurements_tasks.Select(task => task.Result).ToList().ForEach(measurements => u_values.AddRange(measurements));
u_values = u_values.OrderBy(u_value => u_value.value).ToList();
for (int position = 0; position < u_values.Count;)
{
u_values[position].range = ++position;
}
var n1 = number_of_measurements;
var n2 = number_of_measurements;
var R1 = TotalRange(u_values, first_configuration);
var R2 = TotalRange(u_values, second_configuration);
var U1 = Get_U_Statistic(n1, n2, R1);
var U2 = Get_U_Statistic(n2, n1, R2);
var U = Math.Min(U1, U2);
var Z = Math.Round(Math.Abs((U - (float)(n1 * n2) / 2) / Math.Sqrt((float)(n1 * n2 * (n1 + n2 + 1)) / 12)), 2);
Table.Display($"Estadístico Z: {Z} {( Z > 1.96 ? ">" : "<=" )} 1.96 | α: 5%\n");
if (Z > 1.96)
{
return(R1 <= R2 ? first_configuration : second_configuration);
}
return(default);
//Segmenta el hiperespacio alrededor de un punto central de forma hipercúbica para luego procesar concurrentemente cada cluster mediante Black Hole,
//con lo que se obtiene una configuración candidata por cada hipercubo. Finalmente se realiza un torneo asistido por Stateless Q-learning para determinar
//cuál de las configuraciones candidatas es mejor considerando los resultados y desviaciones estandar obtenidas tras una serie de evaluaciones por ronda.
public static async Task <ConfigurationType> Run <ConfigurationType>(ConfigurationType middle_configuration, float hypercube_length, int factor = 1, int depth = 1, int inner_depth = 1, int evaluation_duration = 200, int number_of_epochs = 10, int iterations_by_round = 100, int duration_step_between_rounds = 250, int number_of_performances_to_average = 2, int verification_duration = 1000) where ConfigurationType : IConfiguration
{
var candidate_solutions = await Run_Hypercubic_Black_Hole(middle_configuration, hypercube_length, factor, depth, inner_depth, evaluation_duration, number_of_epochs);
Console.WriteLine($"\nSe han encontrado las siguientes {candidate_solutions.Count} soluciones candidatas luego de aplicar HyperCubic Black Hole:");
for (int index = 0; index < candidate_solutions.Count; index++)
{
Console.WriteLine($"{index + 1}. {candidate_solutions[index]}");
}
Console.WriteLine("\nA continuación se determinará cuál de estas soluciones es mejor mediante un torneo asistido por Stateless Q Learning.\n");
candidate_solutions = candidate_solutions.Distinct().ToList();
var best_configuration = Run_Stateless_Q_Learning(candidate_solutions.ToArray(), iterations_by_round, duration_step_between_rounds, number_of_performances_to_average);
var Table = new DynamicTable(Console.CursorTop, "Comprobación de resultado", new string[] { "Configuración", "IGD" }, candidate_solutions.Select(solution => new string[] { solution.ToString(), "─" }).ToArray());
Table.Display($"Generación final: {verification_duration}");
for (int index = 0; index < candidate_solutions.Count; index++)
{
Table.Update(decimal.Parse(candidate_solutions[index].Evaluate(verification_duration).ToString(), NumberStyles.Any, CultureInfo.InvariantCulture).ToString(), index, 1);
}
return(best_configuration);
}
//Genera una red hipercúbica alrededor de la configuración vectorizada central, para luego aplicar el black hole concurrentemente en cada hipercubo.
public static async Task <List <ConfigurationType> > Run <ConfigurationType>(ConfigurationType middle_configuration, float length, int factor, int depth, int inner_depth, int evaluation_duration, int number_of_epochs) where ConfigurationType : IConfiguration
{
if (middle_configuration == null)
{
throw new ArgumentNullException("<middle_configuration> no puede ser nulo.");
}
var path_to_folder_container = @"C:\Users\Trifenix\Desktop\Tesis_Learnheuristics\HyperSpace";
if (Directory.Exists(path_to_folder_container))
{
Directory.Delete(path_to_folder_container, true);
}
var HyperSpace = HyperCube.CreateHyperCubeNetwork(middle_configuration.Vectorized_Configuration, length, factor, depth, inner_depth);
Console.WriteLine($"A continuación se procesarán los {HyperSpace.Count} hipercubos concurrentemente.\n");
Table = new DynamicTable(Console.CursorTop, "Progreso", new string[] { "HiperCubo", "Época" }, HyperSpace.Select((hypercube, index) => new string[] { (index + 1).ToString(), "0" }).ToArray(), $"Época final: {number_of_epochs}");
Table.Display();
ProblemName = middle_configuration.ProblemName;
var black_holes_hypercubes_task = new List <Task <ConfigurationType> >();
for (int i = 0; i < HyperSpace.Count; i++)
{
var index = i;
var hypercube = HyperSpace[index];
var task = Task.Run(() => ProcessHyperCube <ConfigurationType>(hypercube, $"HyperCube_{++index}", evaluation_duration, number_of_epochs));
black_holes_hypercubes_task.Add(task);
}
// Esperando por las tareas (Un black hole por hipercubo).
await Task.WhenAll(black_holes_hypercubes_task);
var candidate_solutions = new List <ConfigurationType>();
black_holes_hypercubes_task.Select(task => task.Result).Where(result => result != null).ToList().ForEach(candidate_solution => candidate_solutions.Add(candidate_solution));
return(candidate_solutions);
}
