[13852] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
| 3 | * Copyright (C) 2002-2016 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
| 4 | *
|
---|
| 5 | * This file is part of HeuristicLab.
|
---|
| 6 | *
|
---|
| 7 | *
|
---|
| 8 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
| 9 | * it under the terms of the GNU General Public License as published by
|
---|
| 10 | * the Free Software Foundation, either version 3 of the License, or
|
---|
| 11 | * (at your option) any later version.
|
---|
| 12 | *
|
---|
| 13 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
| 16 | * GNU General Public License for more details.
|
---|
| 17 | *
|
---|
| 18 | * You should have received a copy of the GNU General Public License
|
---|
| 19 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
| 20 | */
|
---|
| 21 | #endregion
|
---|
| 22 | using System;
|
---|
[13749] | 23 | using System.Linq;
|
---|
| 24 | using System.Collections.Generic;
|
---|
| 25 | using HeuristicLab.Analysis;
|
---|
| 26 | using HeuristicLab.Common;
|
---|
| 27 | using HeuristicLab.Core;
|
---|
| 28 | using HeuristicLab.Data;
|
---|
| 29 | using HeuristicLab.Encodings.RealVectorEncoding;
|
---|
| 30 | using HeuristicLab.Operators;
|
---|
| 31 | using HeuristicLab.Optimization;
|
---|
| 32 | using HeuristicLab.Optimization.Operators;
|
---|
| 33 | using HeuristicLab.Parameters;
|
---|
| 34 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 35 | using HeuristicLab.PluginInfrastructure;
|
---|
| 36 | using HeuristicLab.Problems.MultiObjectiveTestFunctions;
|
---|
| 37 | using HeuristicLab.Random;
|
---|
| 38 | using System.Threading;
|
---|
[13756] | 39 | using HeuristicLab.Algorithms.GDE3;
|
---|
[13749] | 40 |
|
---|
| 41 | namespace HeuristicLab.Algoritms.GDE3
|
---|
| 42 | {
|
---|
| 43 |
|
---|
| 44 | [Item("Generalized Differential Evolution (GDE3)", "A generalized differential evolution algorithm.")]
|
---|
| 45 | [StorableClass]
|
---|
| 46 | [Creatable(CreatableAttribute.Categories.PopulationBasedAlgorithms, Priority = 400)]
|
---|
[13756] | 47 | public class GDE3 : BasicAlgorithm
|
---|
[13749] | 48 | {
|
---|
| 49 | public override Type ProblemType
|
---|
| 50 | {
|
---|
| 51 | get { return typeof(MultiObjectiveTestFunctionProblem); }
|
---|
| 52 | }
|
---|
[13756] | 53 | public new MultiObjectiveTestFunctionProblem Problem
|
---|
[13749] | 54 | {
|
---|
| 55 | get { return (MultiObjectiveTestFunctionProblem)base.Problem; }
|
---|
| 56 | set { base.Problem = value; }
|
---|
| 57 | }
|
---|
| 58 |
|
---|
| 59 | public ILookupParameter<DoubleMatrix> BestKnownFrontParameter
|
---|
| 60 | {
|
---|
| 61 | get
|
---|
| 62 | {
|
---|
| 63 | return (ILookupParameter<DoubleMatrix>)Parameters["BestKnownFront"];
|
---|
| 64 | }
|
---|
| 65 | }
|
---|
| 66 |
|
---|
| 67 | private readonly IRandom _random = new MersenneTwister();
|
---|
| 68 | private int evals;
|
---|
[13849] | 69 | private double IGDSumm;
|
---|
[13749] | 70 |
|
---|
| 71 | #region ParameterNames
|
---|
[13756] | 72 | private const string MaximumGenerationsParameterName = "Maximum Generations";
|
---|
[13749] | 73 | private const string CrossoverProbabilityParameterName = "CrossoverProbability";
|
---|
| 74 | private const string PopulationSizeParameterName = "PopulationSize";
|
---|
| 75 | private const string ScalingFactorParameterName = "ScalingFactor";
|
---|
[13849] | 76 |
|
---|
[13749] | 77 | #endregion
|
---|
| 78 |
|
---|
| 79 | #region ParameterProperties
|
---|
[13756] | 80 | public IFixedValueParameter<IntValue> MaximumGenerationsParameter
|
---|
[13749] | 81 | {
|
---|
[13756] | 82 | get { return (IFixedValueParameter<IntValue>)Parameters[MaximumGenerationsParameterName]; }
|
---|
[13749] | 83 | }
|
---|
| 84 | private ValueParameter<IntValue> PopulationSizeParameter
|
---|
| 85 | {
|
---|
| 86 | get { return (ValueParameter<IntValue>)Parameters[PopulationSizeParameterName]; }
|
---|
| 87 | }
|
---|
| 88 | public ValueParameter<DoubleValue> CrossoverProbabilityParameter
|
---|
| 89 | {
|
---|
| 90 | get { return (ValueParameter<DoubleValue>)Parameters[CrossoverProbabilityParameterName]; }
|
---|
| 91 | }
|
---|
| 92 | public ValueParameter<DoubleValue> ScalingFactorParameter
|
---|
| 93 | {
|
---|
| 94 | get { return (ValueParameter<DoubleValue>)Parameters[ScalingFactorParameterName]; }
|
---|
| 95 | }
|
---|
| 96 | #endregion
|
---|
| 97 |
|
---|
| 98 | #region Properties
|
---|
| 99 | public int MaximumEvaluations
|
---|
| 100 | {
|
---|
[13756] | 101 | get { return MaximumGenerationsParameter.Value.Value; }
|
---|
| 102 | set { MaximumGenerationsParameter.Value.Value = value; }
|
---|
[13749] | 103 | }
|
---|
| 104 |
|
---|
| 105 | public Double CrossoverProbability
|
---|
| 106 | {
|
---|
| 107 | get { return CrossoverProbabilityParameter.Value.Value; }
|
---|
| 108 | set { CrossoverProbabilityParameter.Value.Value = value; }
|
---|
| 109 | }
|
---|
| 110 | public Double ScalingFactor
|
---|
| 111 | {
|
---|
| 112 | get { return ScalingFactorParameter.Value.Value; }
|
---|
| 113 | set { ScalingFactorParameter.Value.Value = value; }
|
---|
| 114 | }
|
---|
| 115 | public IntValue PopulationSize
|
---|
| 116 | {
|
---|
| 117 | get { return PopulationSizeParameter.Value; }
|
---|
| 118 | set { PopulationSizeParameter.Value = value; }
|
---|
| 119 | }
|
---|
| 120 | #endregion
|
---|
| 121 |
|
---|
| 122 | #region ResultsProperties
|
---|
| 123 | private double ResultsBestQuality
|
---|
| 124 | {
|
---|
| 125 | get { return ((DoubleValue)Results["Best Quality"].Value).Value; }
|
---|
| 126 | set { ((DoubleValue)Results["Best Quality"].Value).Value = value; }
|
---|
| 127 | }
|
---|
| 128 |
|
---|
[13849] | 129 | private double ResultsIGDMean
|
---|
| 130 | {
|
---|
| 131 | get { return ((DoubleValue)Results["IGDMeanValue"].Value).Value; }
|
---|
| 132 | set { ((DoubleValue)Results["IGDMeanValue"].Value).Value = value; }
|
---|
| 133 | }
|
---|
| 134 |
|
---|
| 135 | private double ResultsIGDBest
|
---|
| 136 | {
|
---|
| 137 | get { return ((DoubleValue)Results["IGDBestValue"].Value).Value; }
|
---|
| 138 | set { ((DoubleValue)Results["IGDBestValue"].Value).Value = value; }
|
---|
| 139 | }
|
---|
| 140 |
|
---|
| 141 | private double ResultsIGDWorst
|
---|
| 142 | {
|
---|
| 143 | get { return ((DoubleValue)Results["IGDWorstValue"].Value).Value; }
|
---|
| 144 | set { ((DoubleValue)Results["IGDWorstValue"].Value).Value = value; }
|
---|
| 145 | }
|
---|
| 146 |
|
---|
[13756] | 147 | private double ResultsInvertedGenerationalDistance
|
---|
[13749] | 148 | {
|
---|
| 149 | get { return ((DoubleValue)Results["InvertedGenerationalDistance"].Value).Value; }
|
---|
| 150 | set { ((DoubleValue)Results["InvertedGenerationalDistance"].Value).Value = value; }
|
---|
| 151 | }
|
---|
| 152 |
|
---|
| 153 | private double ResultsHypervolume
|
---|
| 154 | {
|
---|
[13756] | 155 | get { return ((DoubleValue)Results["HyperVolumeValue"].Value).Value; }
|
---|
| 156 | set { ((DoubleValue)Results["HyperVolumeValue"].Value).Value = value; }
|
---|
[13749] | 157 | }
|
---|
| 158 |
|
---|
| 159 | private DoubleMatrix ResultsBestFront
|
---|
| 160 | {
|
---|
| 161 | get { return (DoubleMatrix)Results["Best Front"].Value; }
|
---|
| 162 | set { Results["Best Front"].Value = value; }
|
---|
| 163 | }
|
---|
| 164 |
|
---|
| 165 | private int ResultsEvaluations
|
---|
| 166 | {
|
---|
| 167 | get { return ((IntValue)Results["Evaluations"].Value).Value; }
|
---|
| 168 | set { ((IntValue)Results["Evaluations"].Value).Value = value; }
|
---|
| 169 | }
|
---|
[13756] | 170 | private int ResultsGenerations
|
---|
| 171 | {
|
---|
| 172 | get { return ((IntValue)Results["Generations"].Value).Value; }
|
---|
| 173 | set { ((IntValue)Results["Generations"].Value).Value = value; }
|
---|
| 174 | }
|
---|
| 175 | private double ResultsGenerationalDistance
|
---|
| 176 | {
|
---|
| 177 | get { return ((DoubleValue)Results["GenerationalDistance"].Value).Value; }
|
---|
| 178 | set { ((DoubleValue)Results["GenerationalDistance"].Value).Value = value; }
|
---|
| 179 | }
|
---|
[13749] | 180 |
|
---|
[13756] | 181 | private double ResultsSpacing
|
---|
[13749] | 182 | {
|
---|
[13756] | 183 | get { return ((DoubleValue)Results["Spacing"].Value).Value; }
|
---|
| 184 | set { ((DoubleValue)Results["Spacing"].Value).Value = value; }
|
---|
[13749] | 185 | }
|
---|
[13756] | 186 |
|
---|
| 187 | private double ResultsCrowding
|
---|
[13749] | 188 | {
|
---|
[13756] | 189 | get { return ((DoubleValue)Results["Crowding"].Value).Value; }
|
---|
| 190 | set { ((DoubleValue)Results["Crowding"].Value).Value = value; }
|
---|
[13749] | 191 | }
|
---|
| 192 |
|
---|
| 193 | #endregion
|
---|
| 194 |
|
---|
| 195 | [StorableConstructor]
|
---|
| 196 | protected GDE3(bool deserializing) : base(deserializing) { }
|
---|
| 197 |
|
---|
| 198 | protected GDE3(GDE3 original, Cloner cloner)
|
---|
| 199 | : base(original, cloner)
|
---|
| 200 | {
|
---|
| 201 | }
|
---|
| 202 |
|
---|
| 203 | public override IDeepCloneable Clone(Cloner cloner)
|
---|
| 204 | {
|
---|
| 205 | return new GDE3(this, cloner);
|
---|
| 206 | }
|
---|
| 207 |
|
---|
| 208 | public GDE3()
|
---|
| 209 | {
|
---|
[13756] | 210 | Parameters.Add(new FixedValueParameter<IntValue>(MaximumGenerationsParameterName, "", new IntValue(1000)));
|
---|
[13749] | 211 | Parameters.Add(new ValueParameter<IntValue>(PopulationSizeParameterName, "The size of the population of solutions.", new IntValue(100)));
|
---|
| 212 | Parameters.Add(new ValueParameter<DoubleValue>(CrossoverProbabilityParameterName, "The value for crossover rate", new DoubleValue(0.5)));
|
---|
| 213 | Parameters.Add(new ValueParameter<DoubleValue>(ScalingFactorParameterName, "The value for scaling factor", new DoubleValue(0.5)));
|
---|
| 214 | Parameters.Add(new LookupParameter<DoubleMatrix>("BestKnownFront", "The currently best known Pareto front"));
|
---|
| 215 | }
|
---|
| 216 |
|
---|
| 217 | protected override void Run(CancellationToken cancellationToken)
|
---|
| 218 | {
|
---|
| 219 | // Set up the results display
|
---|
[13756] | 220 | Results.Add(new Result("Generations", new IntValue(0)));
|
---|
[13749] | 221 | Results.Add(new Result("Evaluations", new IntValue(0)));
|
---|
| 222 | Results.Add(new Result("Best Front", new DoubleMatrix()));
|
---|
[13756] | 223 | Results.Add(new Result("Crowding", new DoubleValue(0)));
|
---|
[13749] | 224 | Results.Add(new Result("InvertedGenerationalDistance", new DoubleValue(0)));
|
---|
[13756] | 225 | Results.Add(new Result("GenerationalDistance", new DoubleValue(0)));
|
---|
| 226 | Results.Add(new Result("HyperVolumeValue", new DoubleValue(0)));
|
---|
[13849] | 227 | Results.Add(new Result("IGDMeanValue", new DoubleValue(0)));
|
---|
| 228 | Results.Add(new Result("IGDBestValue", new DoubleValue(Int32.MaxValue)));
|
---|
| 229 | Results.Add(new Result("IGDWorstValue", new DoubleValue(0)));
|
---|
| 230 |
|
---|
[13756] | 231 | Results.Add(new Result("Spacing", new DoubleValue(0)));
|
---|
[13749] | 232 | Results.Add(new Result("Scatterplot", typeof(IMOFrontModel)));
|
---|
| 233 | var table = new DataTable("Qualities");
|
---|
| 234 | table.Rows.Add(new DataRow("Best Quality"));
|
---|
| 235 | Results.Add(new Result("Qualities", table));
|
---|
| 236 |
|
---|
[13756] | 237 | //setup the variables
|
---|
| 238 | List<SolutionSet> population;
|
---|
| 239 | List<SolutionSet> offspringPopulation;
|
---|
| 240 | SolutionSet[] parent;
|
---|
[13849] | 241 | double IGDSumm = 0;
|
---|
[13756] | 242 |
|
---|
| 243 | //initialize population
|
---|
| 244 | population = new List<SolutionSet>(PopulationSizeParameter.Value.Value);
|
---|
[13749] | 245 |
|
---|
| 246 | for (int i = 0; i < PopulationSizeParameter.Value.Value; ++i)
|
---|
| 247 | {
|
---|
| 248 | var m = createIndividual();
|
---|
[13849] | 249 | m.Quality = Problem.Evaluate(m.Population, _random);
|
---|
| 250 | //the test function is constrained
|
---|
| 251 | if (m.Quality.Length > Problem.Objectives)
|
---|
| 252 | {
|
---|
| 253 | m.OverallConstrainViolation = m.Quality[Problem.Objectives];
|
---|
| 254 | } else {
|
---|
| 255 | m.OverallConstrainViolation = 0;
|
---|
| 256 | }
|
---|
[13749] | 257 | population.Add(m);
|
---|
| 258 | }
|
---|
| 259 |
|
---|
| 260 | this.initProgress();
|
---|
[13849] | 261 | int generations = 1;
|
---|
[13749] | 262 |
|
---|
[13756] | 263 | while (ResultsGenerations < MaximumGenerationsParameter.Value.Value
|
---|
[13749] | 264 | && !cancellationToken.IsCancellationRequested)
|
---|
| 265 | {
|
---|
[13756] | 266 | var populationSize = PopulationSizeParameter.Value.Value;
|
---|
[13749] | 267 |
|
---|
[13756] | 268 | // Create the offSpring solutionSet
|
---|
| 269 | offspringPopulation = new List<SolutionSet>(PopulationSizeParameter.Value.Value * 2);
|
---|
[13749] | 270 |
|
---|
[13756] | 271 | for (int i = 0; i < populationSize; i++)
|
---|
| 272 | {
|
---|
| 273 | // Obtain parents. Two parameters are required: the population and the
|
---|
| 274 | // index of the current individual
|
---|
| 275 | parent = selection(population, i);
|
---|
[13749] | 276 |
|
---|
[13756] | 277 | SolutionSet child;
|
---|
| 278 | // Crossover. The parameters are the current individual and the index of the array of parents
|
---|
| 279 | child = reproduction(population[i], parent);
|
---|
| 280 |
|
---|
| 281 | child.Quality = Problem.Evaluate(child.Population, _random);
|
---|
[13849] | 282 |
|
---|
[13756] | 283 | this.updateProgres();
|
---|
| 284 |
|
---|
[13849] | 285 | //the test function is constrained
|
---|
| 286 | if (child.Quality.Length > Problem.Objectives)
|
---|
| 287 | {
|
---|
| 288 | child.OverallConstrainViolation = child.Quality[Problem.Objectives];
|
---|
| 289 | } else {
|
---|
| 290 | child.OverallConstrainViolation = 0;
|
---|
| 291 | }
|
---|
| 292 |
|
---|
[13756] | 293 | // Dominance test
|
---|
| 294 | int result;
|
---|
[13849] | 295 | result = compareDomination(population[i], child);
|
---|
| 296 |
|
---|
[13756] | 297 | if (result == -1)
|
---|
| 298 | { // Solution i dominates child
|
---|
| 299 | offspringPopulation.Add(population[i]);
|
---|
[13749] | 300 | }
|
---|
[13756] | 301 | else if (result == 1)
|
---|
| 302 | { // child dominates
|
---|
| 303 | offspringPopulation.Add(child);
|
---|
| 304 | }
|
---|
| 305 | else
|
---|
| 306 | { // the two solutions are non-dominated
|
---|
| 307 | offspringPopulation.Add(child);
|
---|
| 308 | offspringPopulation.Add(population[i]);
|
---|
| 309 | }
|
---|
[13749] | 310 | }
|
---|
[13849] | 311 |
|
---|
[13756] | 312 | // Ranking the offspring population
|
---|
| 313 | List<SolutionSet>[] ranking = computeRanking(offspringPopulation);
|
---|
[13849] | 314 | population = crowdingDistanceSelection(ranking);
|
---|
| 315 | generations++;
|
---|
| 316 | ResultsGenerations = generations;
|
---|
| 317 | displayResults(population);
|
---|
[13756] | 318 | }
|
---|
| 319 | }
|
---|
[13749] | 320 |
|
---|
[13756] | 321 | private void displayResults(List<SolutionSet> population)
|
---|
| 322 | {
|
---|
| 323 | List<SolutionSet>[] rankingFinal = computeRanking(population);
|
---|
| 324 |
|
---|
| 325 | int objectives = Problem.Objectives;
|
---|
| 326 | var optimalfront = Problem.TestFunction.OptimalParetoFront(objectives);
|
---|
| 327 |
|
---|
| 328 | double[][] opf = new double[0][];
|
---|
| 329 | if (optimalfront != null)
|
---|
| 330 | {
|
---|
| 331 | opf = optimalfront.Select(s => s.ToArray()).ToArray();
|
---|
| 332 | }
|
---|
| 333 |
|
---|
[13849] | 334 | //compute the final qualities and population
|
---|
[13756] | 335 | double[][] qualitiesFinal = new double[rankingFinal[0].Count][];
|
---|
[13849] | 336 | double[][] populationFinal = new double[rankingFinal[0].Count][];
|
---|
[13756] | 337 |
|
---|
| 338 | for (int i = 0; i < rankingFinal[0].Count; ++i)
|
---|
| 339 | {
|
---|
| 340 | qualitiesFinal[i] = new double[Problem.Objectives];
|
---|
[13849] | 341 | populationFinal[i] = new double[Problem.Objectives];
|
---|
[13756] | 342 | for (int j = 0; j < Problem.Objectives; ++j)
|
---|
[13749] | 343 | {
|
---|
[13849] | 344 | populationFinal[i][j] = rankingFinal[0][i].Population[j];
|
---|
[13756] | 345 | qualitiesFinal[i][j] = rankingFinal[0][i].Quality[j];
|
---|
[13749] | 346 | }
|
---|
[13756] | 347 | }
|
---|
| 348 | IEnumerable<double[]> en = qualitiesFinal;
|
---|
| 349 | IEnumerable<double[]> frontVectors = NonDominatedSelect.selectNonDominatedVectors(qualitiesFinal, Problem.TestFunction.Maximization(objectives), true);
|
---|
| 350 | //update the results
|
---|
[13749] | 351 |
|
---|
[13756] | 352 | ResultsEvaluations = this.evals;
|
---|
| 353 | ResultsBestFront = new DoubleMatrix(MultiObjectiveTestFunctionProblem.To2D(qualitiesFinal));
|
---|
| 354 | ResultsCrowding = Crowding.Calculate(qualitiesFinal, Problem.TestFunction.Bounds(objectives));
|
---|
| 355 | ResultsInvertedGenerationalDistance = InvertedGenerationalDistance.Calculate(en, optimalfront, 1);
|
---|
| 356 | ResultsHypervolume = Hypervolume.Calculate(frontVectors, Problem.TestFunction.ReferencePoint(objectives), Problem.TestFunction.Maximization(objectives));
|
---|
| 357 | ResultsGenerationalDistance = GenerationalDistance.Calculate(qualitiesFinal, optimalfront, 1);
|
---|
| 358 | Results["Scatterplot"].Value = new MOSolution(qualitiesFinal, populationFinal, opf, objectives);
|
---|
| 359 | ResultsSpacing = Spacing.Calculate(qualitiesFinal);
|
---|
[13849] | 360 |
|
---|
| 361 | if (ResultsIGDBest > ResultsInvertedGenerationalDistance) {
|
---|
| 362 | ResultsIGDBest = ResultsInvertedGenerationalDistance;
|
---|
| 363 | }
|
---|
| 364 | if (ResultsIGDWorst < ResultsInvertedGenerationalDistance)
|
---|
| 365 | {
|
---|
| 366 | ResultsIGDWorst = ResultsInvertedGenerationalDistance;
|
---|
| 367 | }
|
---|
| 368 | this.IGDSumm += ResultsInvertedGenerationalDistance;
|
---|
| 369 | ResultsIGDMean = this.IGDSumm / ResultsGenerations;
|
---|
[13756] | 370 | }
|
---|
| 371 |
|
---|
| 372 | private int getWorstIndex(List<SolutionSet> SolutionsList)
|
---|
| 373 | {
|
---|
| 374 | int result = 0;
|
---|
| 375 |
|
---|
| 376 | if ((SolutionsList == null) || SolutionsList.Count == 0)
|
---|
| 377 | {
|
---|
| 378 | result = 0;
|
---|
[13749] | 379 | }
|
---|
[13756] | 380 | else
|
---|
| 381 | {
|
---|
| 382 | SolutionSet worstKnown = SolutionsList[0],
|
---|
| 383 | candidateSolution;
|
---|
| 384 | int flag;
|
---|
| 385 | for (int i = 1; i < SolutionsList.Count; i++)
|
---|
| 386 | {
|
---|
| 387 | candidateSolution = SolutionsList[i];
|
---|
[13849] | 388 | flag = compareDomination(worstKnown, candidateSolution);
|
---|
[13756] | 389 | if (flag == -1)
|
---|
| 390 | {
|
---|
| 391 | result = i;
|
---|
| 392 | worstKnown = candidateSolution;
|
---|
| 393 | }
|
---|
| 394 | }
|
---|
| 395 | }
|
---|
| 396 | return result;
|
---|
[13749] | 397 | }
|
---|
[13756] | 398 |
|
---|
[13749] | 399 | protected SolutionSet createIndividual()
|
---|
| 400 | {
|
---|
| 401 | var dim = Problem.ProblemSize;
|
---|
| 402 | var lb = Problem.Bounds[0, 0];
|
---|
| 403 | var ub = Problem.Bounds[0, 1];
|
---|
| 404 | var range = ub - lb;
|
---|
| 405 | var v = new double[Problem.ProblemSize];
|
---|
| 406 | SolutionSet solutionObject = new SolutionSet(PopulationSizeParameter.Value.Value);
|
---|
| 407 |
|
---|
| 408 | for (int i = 0; i < Problem.ProblemSize; ++i)
|
---|
| 409 | {
|
---|
| 410 | v[i] = _random.NextDouble() * range + lb;
|
---|
| 411 |
|
---|
| 412 | }
|
---|
[13849] | 413 | solutionObject.createSolution(v);
|
---|
[13749] | 414 | return solutionObject;
|
---|
| 415 | }
|
---|
| 416 |
|
---|
| 417 | private SolutionSet createEmptyIndividual()
|
---|
| 418 | {
|
---|
| 419 | SolutionSet solutionObject = new SolutionSet(PopulationSizeParameter.Value.Value);
|
---|
| 420 | var n = new RealVector(Problem.ProblemSize);
|
---|
| 421 | solutionObject.Population = n;
|
---|
| 422 | return solutionObject;
|
---|
| 423 | }
|
---|
| 424 |
|
---|
| 425 | protected void initProgress()
|
---|
| 426 | {
|
---|
| 427 | this.evals = PopulationSizeParameter.Value.Value;
|
---|
| 428 | }
|
---|
| 429 |
|
---|
| 430 | protected void updateProgres()
|
---|
| 431 | {
|
---|
[13756] | 432 | this.evals++;
|
---|
[13749] | 433 | }
|
---|
| 434 |
|
---|
[13756] | 435 | protected SolutionSet[] selection(List<SolutionSet> population, int i)
|
---|
[13749] | 436 | {
|
---|
[13756] | 437 | SolutionSet[] parents = new SolutionSet[3];
|
---|
| 438 | int r0, r1, r2;
|
---|
| 439 | //assure the selected vectors r0, r1 and r2 are different
|
---|
| 440 | do
|
---|
[13749] | 441 | {
|
---|
[13756] | 442 | r0 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
| 443 | } while (r0 == i);
|
---|
| 444 | do
|
---|
| 445 | {
|
---|
| 446 | r1 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
| 447 | } while (r1 == i || r1 == r0);
|
---|
| 448 | do
|
---|
| 449 | {
|
---|
| 450 | r2 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
| 451 | } while (r2 == i || r2 == r0 || r2 == r1);
|
---|
[13749] | 452 |
|
---|
[13756] | 453 | parents[0] = population[r0];
|
---|
| 454 | parents[1] = population[r1];
|
---|
| 455 | parents[2] = population[r2];
|
---|
| 456 |
|
---|
[13749] | 457 | return parents;
|
---|
| 458 | }
|
---|
| 459 |
|
---|
[13756] | 460 | protected SolutionSet reproduction(SolutionSet parent, SolutionSet[] parentsSolutions)
|
---|
[13749] | 461 | {
|
---|
[13756] | 462 | var individual = createEmptyIndividual();
|
---|
| 463 | double rnbr = _random.Next(0, Problem.ProblemSize);
|
---|
| 464 | for (int m = 0; m < Problem.ProblemSize; m++)
|
---|
[13749] | 465 | {
|
---|
[13756] | 466 | if (_random.NextDouble() < CrossoverProbabilityParameter.Value.Value || m == rnbr)
|
---|
| 467 | {
|
---|
| 468 | double value;
|
---|
| 469 | value = parentsSolutions[2].Population[m] +
|
---|
| 470 | ScalingFactorParameter.Value.Value * (parentsSolutions[0].Population[m] - parentsSolutions[1].Population[m]);
|
---|
| 471 | //check the problem upper and lower bounds
|
---|
| 472 | if (value > Problem.Bounds[0, 1]) value = Problem.Bounds[0, 1];
|
---|
| 473 | if (value < Problem.Bounds[0, 0]) value = Problem.Bounds[0, 0];
|
---|
| 474 | individual.Population[m] = value;
|
---|
[13749] | 475 | }
|
---|
[13756] | 476 | else
|
---|
[13749] | 477 | {
|
---|
[13756] | 478 | double value;
|
---|
| 479 | value = parent.Population[m];
|
---|
| 480 | individual.Population[m] = value;
|
---|
[13749] | 481 | }
|
---|
| 482 | }
|
---|
[13756] | 483 | return individual;
|
---|
[13749] | 484 | }
|
---|
| 485 |
|
---|
| 486 | private List<SolutionSet> crowdingDistanceSelection(List<SolutionSet>[] ranking)
|
---|
| 487 | {
|
---|
| 488 | List<SolutionSet> population = new List<SolutionSet>();
|
---|
| 489 | int rankingIndex = 0;
|
---|
| 490 | while (populationIsNotFull(population))
|
---|
| 491 | {
|
---|
[13756] | 492 | if (subFrontFillsIntoThePopulation(ranking, rankingIndex, population))
|
---|
| 493 | {
|
---|
[13749] | 494 | addRankedSolutionToPopulation(ranking, rankingIndex, population);
|
---|
| 495 | rankingIndex++;
|
---|
[13756] | 496 | }
|
---|
| 497 | else {
|
---|
[13849] | 498 | crowdingDistanceAssignment(ranking[rankingIndex]);
|
---|
[13749] | 499 | addLastRankedSolutionToPopulation(ranking, rankingIndex, population);
|
---|
| 500 | }
|
---|
| 501 | }
|
---|
| 502 | return population;
|
---|
| 503 | }
|
---|
| 504 |
|
---|
| 505 | private void addLastRankedSolutionToPopulation(List<SolutionSet>[] ranking, int rankingIndex, List<SolutionSet> population)
|
---|
| 506 | {
|
---|
| 507 | List<SolutionSet> currentRankedFront = ranking[rankingIndex];
|
---|
[13849] | 508 | //descending sort and add the front with highest crowding distance to the population
|
---|
| 509 | currentRankedFront.Sort((x, y) => -x.CrowdingDistance.CompareTo(y.CrowdingDistance));
|
---|
[13749] | 510 | int i = 0;
|
---|
| 511 | while (population.Count < PopulationSizeParameter.Value.Value)
|
---|
| 512 | {
|
---|
| 513 | population.Add(currentRankedFront[i]);
|
---|
| 514 | i++;
|
---|
| 515 | }
|
---|
| 516 | }
|
---|
| 517 |
|
---|
[13849] | 518 | public void crowdingDistanceAssignment(List<SolutionSet> rankingSubfront)
|
---|
[13749] | 519 | {
|
---|
| 520 | int size = rankingSubfront.Count;
|
---|
| 521 |
|
---|
| 522 | if (size == 0)
|
---|
| 523 | return;
|
---|
| 524 |
|
---|
| 525 | if (size == 1)
|
---|
| 526 | {
|
---|
| 527 | rankingSubfront[0].CrowdingDistance = double.PositiveInfinity;
|
---|
| 528 | return;
|
---|
| 529 | }
|
---|
| 530 |
|
---|
| 531 | if (size == 2)
|
---|
| 532 | {
|
---|
| 533 | rankingSubfront[0].CrowdingDistance = double.PositiveInfinity;
|
---|
| 534 | rankingSubfront[1].CrowdingDistance = double.PositiveInfinity;
|
---|
| 535 | return;
|
---|
| 536 | }
|
---|
| 537 |
|
---|
| 538 | //Use a new SolutionSet to evite alter original solutionSet
|
---|
| 539 | List<SolutionSet> front = new List<SolutionSet>(size);
|
---|
| 540 | for (int i = 0; i < size; i++)
|
---|
| 541 | {
|
---|
| 542 | front.Add(rankingSubfront[i]);
|
---|
| 543 | }
|
---|
| 544 |
|
---|
| 545 | for (int i = 0; i < size; i++)
|
---|
[13849] | 546 | front[i].CrowdingDistance = 0.0;
|
---|
[13749] | 547 |
|
---|
| 548 | double objetiveMaxn;
|
---|
| 549 | double objetiveMinn;
|
---|
| 550 | double distance;
|
---|
| 551 |
|
---|
[13756] | 552 | for (int i = 0; i < Problem.Objectives; i++)
|
---|
[13749] | 553 | {
|
---|
[13849] | 554 | // Sort the front population by the objective i
|
---|
[13749] | 555 | front.Sort((x, y) => x.Quality[i].CompareTo(y.Quality[i]));
|
---|
| 556 | objetiveMinn = front[0].Quality[i];
|
---|
| 557 | objetiveMaxn = front[front.Count - 1].Quality[i];
|
---|
| 558 |
|
---|
[13849] | 559 | //Set crowding distance for the current front
|
---|
[13756] | 560 | front[0].CrowdingDistance = double.PositiveInfinity;
|
---|
[13749] | 561 | front[size - 1].CrowdingDistance = double.PositiveInfinity;
|
---|
| 562 |
|
---|
| 563 | for (int j = 1; j < size - 1; j++)
|
---|
| 564 | {
|
---|
| 565 | distance = front[j + 1].Quality[i] - front[j - 1].Quality[i];
|
---|
| 566 | distance = distance / (objetiveMaxn - objetiveMinn);
|
---|
| 567 | distance += front[j].CrowdingDistance;
|
---|
| 568 | front[j].CrowdingDistance = distance;
|
---|
| 569 | }
|
---|
| 570 | }
|
---|
| 571 | }
|
---|
| 572 |
|
---|
| 573 | private void addRankedSolutionToPopulation(List<SolutionSet>[] ranking, int rankingIndex, List<SolutionSet> population)
|
---|
| 574 | {
|
---|
[13756] | 575 | foreach (SolutionSet solution in ranking[rankingIndex])
|
---|
[13749] | 576 | {
|
---|
| 577 | population.Add(solution);
|
---|
| 578 | }
|
---|
| 579 | }
|
---|
| 580 |
|
---|
| 581 | private bool subFrontFillsIntoThePopulation(List<SolutionSet>[] ranking, int rankingIndex, List<SolutionSet> population)
|
---|
| 582 | {
|
---|
| 583 | return ranking[rankingIndex].Count < (PopulationSizeParameter.Value.Value - population.Count);
|
---|
| 584 | }
|
---|
| 585 |
|
---|
| 586 | private bool populationIsNotFull(List<SolutionSet> population)
|
---|
| 587 | {
|
---|
| 588 | return population.Count < PopulationSizeParameter.Value.Value;
|
---|
| 589 | }
|
---|
| 590 |
|
---|
| 591 | private List<SolutionSet>[] computeRanking(List<SolutionSet> tmpList)
|
---|
| 592 | {
|
---|
| 593 | // dominateMe[i] contains the number of solutions dominating i
|
---|
| 594 | int[] dominateMe = new int[tmpList.Count];
|
---|
| 595 |
|
---|
| 596 | // iDominate[k] contains the list of solutions dominated by k
|
---|
| 597 | List<int>[] iDominate = new List<int>[tmpList.Count];
|
---|
| 598 |
|
---|
| 599 | // front[i] contains the list of individuals belonging to the front i
|
---|
| 600 | List<int>[] front = new List<int>[tmpList.Count + 1];
|
---|
| 601 |
|
---|
| 602 | // flagDominate is an auxiliar encodings.variable
|
---|
| 603 | int flagDominate;
|
---|
| 604 |
|
---|
| 605 | // Initialize the fronts
|
---|
| 606 | for (int i = 0; i < front.Length; i++)
|
---|
| 607 | {
|
---|
| 608 | front[i] = new List<int>();
|
---|
| 609 | }
|
---|
| 610 |
|
---|
| 611 | //-> Fast non dominated sorting algorithm
|
---|
| 612 | // Contribution of Guillaume Jacquenot
|
---|
| 613 | for (int p = 0; p < tmpList.Count; p++)
|
---|
| 614 | {
|
---|
| 615 | // Initialize the list of individuals that i dominate and the number
|
---|
| 616 | // of individuals that dominate me
|
---|
[13849] | 617 | iDominate[p] = new List<int>();
|
---|
[13749] | 618 | dominateMe[p] = 0;
|
---|
| 619 | }
|
---|
| 620 | for (int p = 0; p < (tmpList.Count - 1); p++)
|
---|
| 621 | {
|
---|
| 622 | // For all q individuals , calculate if p dominates q or vice versa
|
---|
| 623 | for (int q = p + 1; q < tmpList.Count; q++)
|
---|
| 624 | {
|
---|
[13849] | 625 | flagDominate = compareConstraintsViolation(tmpList[p], tmpList[q]);
|
---|
| 626 | if (flagDominate == 0) {
|
---|
| 627 | flagDominate = compareDomination(tmpList[p], tmpList[q]);
|
---|
| 628 | }
|
---|
[13749] | 629 | if (flagDominate == -1)
|
---|
| 630 | {
|
---|
| 631 | iDominate[p].Add(q);
|
---|
| 632 | dominateMe[q]++;
|
---|
| 633 | }
|
---|
| 634 | else if (flagDominate == 1)
|
---|
| 635 | {
|
---|
| 636 | iDominate[q].Add(p);
|
---|
| 637 | dominateMe[p]++;
|
---|
| 638 | }
|
---|
| 639 | }
|
---|
| 640 | // If nobody dominates p, p belongs to the first front
|
---|
| 641 | }
|
---|
| 642 | for (int i = 0; i < tmpList.Count; i++)
|
---|
| 643 | {
|
---|
| 644 | if (dominateMe[i] == 0)
|
---|
| 645 | {
|
---|
| 646 | front[0].Add(i);
|
---|
| 647 | tmpList[i].Rank = 0;
|
---|
| 648 | }
|
---|
| 649 | }
|
---|
| 650 |
|
---|
| 651 | //Obtain the rest of fronts
|
---|
| 652 | int k = 0;
|
---|
| 653 |
|
---|
| 654 | while (front[k].Count != 0)
|
---|
| 655 | {
|
---|
| 656 | k++;
|
---|
| 657 | foreach (var it1 in front[k - 1])
|
---|
| 658 | {
|
---|
| 659 | foreach (var it2 in iDominate[it1])
|
---|
| 660 | {
|
---|
| 661 | int index = it2;
|
---|
| 662 | dominateMe[index]--;
|
---|
| 663 | if (dominateMe[index] == 0)
|
---|
| 664 | {
|
---|
| 665 | front[k].Add(index);
|
---|
| 666 | tmpList[index].Rank = k;
|
---|
| 667 | }
|
---|
| 668 | }
|
---|
| 669 | }
|
---|
| 670 | }
|
---|
| 671 | //<-
|
---|
| 672 |
|
---|
| 673 | var rankedSubpopulation = new List<SolutionSet>[k];
|
---|
| 674 | //0,1,2,....,i-1 are front, then i fronts
|
---|
| 675 | for (int j = 0; j < k; j++)
|
---|
| 676 | {
|
---|
| 677 | rankedSubpopulation[j] = new List<SolutionSet>(front[j].Count);
|
---|
| 678 | foreach (var it1 in front[j])
|
---|
| 679 | {
|
---|
| 680 | rankedSubpopulation[j].Add(tmpList[it1]);
|
---|
| 681 | }
|
---|
| 682 | }
|
---|
| 683 | return rankedSubpopulation;
|
---|
| 684 | }
|
---|
| 685 |
|
---|
[13849] | 686 | private int compareDomination(SolutionSet solution1, SolutionSet solution2)
|
---|
[13749] | 687 | {
|
---|
| 688 | int dominate1; // dominate1 indicates if some objective of solution1
|
---|
| 689 | // dominates the same objective in solution2. dominate2
|
---|
| 690 | int dominate2; // is the complementary of dominate1.
|
---|
| 691 |
|
---|
| 692 | dominate1 = 0;
|
---|
| 693 | dominate2 = 0;
|
---|
| 694 |
|
---|
| 695 | int flag; //stores the result of the comparison
|
---|
| 696 |
|
---|
[13849] | 697 | // Test to determine whether at least a solution violates some constraint
|
---|
| 698 | if (needToCompareViolations(solution1, solution2))
|
---|
| 699 | {
|
---|
| 700 | return compareConstraintsViolation(solution1, solution2);
|
---|
| 701 | }
|
---|
| 702 |
|
---|
[13749] | 703 | // Equal number of violated constraints. Applying a dominance Test then
|
---|
| 704 | double value1, value2;
|
---|
| 705 | for (int i = 0; i < Problem.Objectives; i++)
|
---|
| 706 | {
|
---|
[13849] | 707 | value1 = solution1.Quality[i];
|
---|
| 708 | value2 = solution2.Quality[i];
|
---|
[13749] | 709 | if (value1 < value2)
|
---|
| 710 | {
|
---|
| 711 | flag = -1;
|
---|
| 712 | }
|
---|
[13849] | 713 | else if (value2 < value1)
|
---|
[13749] | 714 | {
|
---|
| 715 | flag = 1;
|
---|
| 716 | }
|
---|
| 717 | else
|
---|
| 718 | {
|
---|
| 719 | flag = 0;
|
---|
| 720 | }
|
---|
| 721 |
|
---|
| 722 | if (flag == -1)
|
---|
| 723 | {
|
---|
| 724 | dominate1 = 1;
|
---|
| 725 | }
|
---|
| 726 |
|
---|
| 727 | if (flag == 1)
|
---|
| 728 | {
|
---|
| 729 | dominate2 = 1;
|
---|
| 730 | }
|
---|
| 731 | }
|
---|
| 732 |
|
---|
| 733 | if (dominate1 == dominate2)
|
---|
| 734 | {
|
---|
| 735 | return 0; //No one dominate the other
|
---|
| 736 | }
|
---|
| 737 | if (dominate1 == 1)
|
---|
| 738 | {
|
---|
| 739 | return -1; // solution1 dominate
|
---|
| 740 | }
|
---|
| 741 | return 1; // solution2 dominate
|
---|
| 742 | }
|
---|
[13849] | 743 |
|
---|
| 744 | private bool needToCompareViolations(SolutionSet solution1, SolutionSet solution2)
|
---|
| 745 | {
|
---|
| 746 | bool needToCompare;
|
---|
| 747 | needToCompare = (solution1.OverallConstrainViolation < 0) || (solution2.OverallConstrainViolation < 0);
|
---|
| 748 |
|
---|
| 749 | return needToCompare;
|
---|
| 750 | }
|
---|
| 751 |
|
---|
| 752 | private int compareConstraintsViolation(SolutionSet solution1, SolutionSet solution2)
|
---|
| 753 | {
|
---|
| 754 | int result;
|
---|
| 755 | double overall1, overall2;
|
---|
| 756 | overall1 = solution1.OverallConstrainViolation;
|
---|
| 757 | overall2 = solution2.OverallConstrainViolation;
|
---|
| 758 |
|
---|
| 759 | if ((overall1 < 0) && (overall2 < 0))
|
---|
| 760 | {
|
---|
| 761 | if (overall1 > overall2)
|
---|
| 762 | {
|
---|
| 763 | result = -1;
|
---|
| 764 | }
|
---|
| 765 | else if (overall2 > overall1)
|
---|
| 766 | {
|
---|
| 767 | result = 1;
|
---|
| 768 | }
|
---|
| 769 | else
|
---|
| 770 | {
|
---|
| 771 | result = 0;
|
---|
| 772 | }
|
---|
| 773 | }
|
---|
| 774 | else if ((overall1 == 0) && (overall2 < 0))
|
---|
| 775 | {
|
---|
| 776 | result = -1;
|
---|
| 777 | }
|
---|
| 778 | else if ((overall1 < 0) && (overall2 == 0))
|
---|
| 779 | {
|
---|
| 780 | result = 1;
|
---|
| 781 | }
|
---|
| 782 | else
|
---|
| 783 | {
|
---|
| 784 | result = 0;
|
---|
| 785 | }
|
---|
| 786 | return result;
|
---|
| 787 | }
|
---|
[13749] | 788 | }
|
---|
| 789 | }
|
---|
| 790 |
|
---|
| 791 |
|
---|
| 792 |
|
---|