Free cookie consent management tool by TermsFeed Policy Generator

Changeset 14404 for branches


Ignore:
Timestamp:
11/21/16 15:52:53 (7 years ago)
Author:
bwerth
Message:

#2592 several fixes and cleanups to adapt a more HeuristicLab-Code-Style + renaming of folders and Plugin

Location:
branches
Files:
7 added
6 deleted
8 edited
2 copied
1 moved

Legend:

Unmodified
Added
Removed

  • branches

    • Property svn:global-ignores set to
      bin

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/CrowdingIndicator.cs

    r14269 r14404  
    2020#endregion
    2121
    22 using HeuristicLab.Data;
    2322using System;
     23using System.Collections.Generic;
    2424using System.Linq;
     25using HeuristicLab.Common;
     26using HeuristicLab.Core;
    2527using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
    26 using HeuristicLab.Core;
    27 using HeuristicLab.Common;
    2828using HeuristicLab.Problems.TestFunctions.MultiObjective;
    2929
    … …  
    3131  [Item("CrowdingIndicator", "Selection of Offspring based on CrowdingDistance")]
    3232  [StorableClass]
    33   class CrowdingIndicator : Item, IIndicator {
     33  internal class CrowdingIndicator : Item, IIndicator {
    3434
     35    public int LeastContributer<TR>(IEnumerable<TR> front, Func<TR, double[]> extractor, MultiObjectiveTestFunctionProblem problem) {
     36      var bounds = problem.Bounds;
     37      var extracted = front.Select(extractor.Invoke).ToArray();
     38      var pointsums = new double[extracted.Length];
    3539
    36     public int LeastContributer<R>(R[] front, Func<R, double[]> extractor, MultiObjectiveTestFunctionProblem problem) {
    37       DoubleMatrix bounds = problem.Bounds;
    38       int mindex = 1;
    39       double min = Double.MaxValue;
    40       double[] pointsums = new double[front.Count()];
    41 
    42       for (int dim = 0; dim < problem.Objectives; dim++) {
    43         double[] arr = front.Select(x => extractor.Invoke(x)[dim]).ToArray();
     40      for (var dim = 0; dim < problem.Objectives; dim++) {
     41        var arr = extracted.Select(x => x[dim]).ToArray();
    4442        Array.Sort(arr);
    45         double fmax = problem.Bounds[dim % bounds.Rows, 1];
    46         double fmin = bounds[dim % bounds.Rows, 0];
    47         int pointIdx = 0;
    48         foreach (R point in front) {
    49           double d = 0;
    50           int pos = Array.BinarySearch(arr, extractor.Invoke(point)[dim]);
    51 
    52           d = pos != 0 && pos != arr.Count() - 1 ? (arr[pos + 1] - arr[pos - 1]) / (fmax - fmin) : Double.PositiveInfinity;
     43        var fmax = problem.Bounds[dim % bounds.Rows, 1];
     44        var fmin = bounds[dim % bounds.Rows, 0];
     45        var pointIdx = 0;
     46        foreach (var point in extracted) {
     47          var pos = Array.BinarySearch(arr, point[dim]);
     48          var d = pos != 0 && pos != arr.Length - 1 ? (arr[pos + 1] - arr[pos - 1]) / (fmax - fmin) : double.PositiveInfinity;
    5349          pointsums[pointIdx] += d;
    54 
    55 
    5650          pointIdx++;
    5751        }
    5852      }
    5953      //find min
    60       for (int pointIdx = 0; pointIdx < pointsums.Length; pointIdx++) {
    61         double d = pointsums[pointIdx];
    62         if (!Double.IsInfinity(d) && min > d) {
    63           mindex = pointIdx;
    64           min = d;
    65         }
    66       }
    67 
    68       return mindex;
    69     }
    70 
    71 
    72 
    73     public override IDeepCloneable Clone(Cloner cloner) {
    74       return new CrowdingIndicator(this, cloner);
     54      return pointsums.Where(x => !double.IsInfinity(x)).ArgMin();
    7555    }
    7656
    7757    [StorableConstructor]
    7858    protected CrowdingIndicator(bool deserializing) : base(deserializing) { }
    79     protected CrowdingIndicator(CrowdingIndicator original, Cloner cloner)
    80       : base(original, cloner) {
    81     }
     59    protected CrowdingIndicator(CrowdingIndicator original, Cloner cloner) : base(original, cloner) { }
     60    public override IDeepCloneable Clone(Cloner cloner) { return new CrowdingIndicator(this, cloner); }
    8261    public CrowdingIndicator() { }
    8362  }

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/HypervolumeIndicator.cs

    r14269 r14404  
    2626using HeuristicLab.Common;
    2727using HeuristicLab.Core;
    28 using HeuristicLab.Data;
    2928using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
    3029using HeuristicLab.Problems.TestFunctions.MultiObjective;
    … …  
    3231  [Item("HypervolumeIndicator", "Selection of Offspring based on contributing Hypervolume")]
    3332  [StorableClass]
    34   class HypervolumeIndicator : Item, IIndicator {
     33  internal class HypervolumeIndicator : Item, IIndicator {
    3534
    36     public int LeastContributer<R>(R[] front, Func<R, double[]> extractor, MultiObjectiveTestFunctionProblem problem) {
    37       DoubleMatrix bounds = problem.Bounds;
    38       int mindex = 1;
    39       double min = Double.MaxValue;
    40       List<double[]> frontArr = front.Select(x => extractor.Invoke(x)).ToList<double[]>();
    41       double[] refPoint = buildReference(frontArr, problem.Maximization);
    42       refPoint = buildReference(refPoint, problem.TestFunction.ReferencePoint(problem.Objectives), problem.Maximization);
    43       double hv = Hypervolume.Calculate(frontArr, refPoint, problem.Maximization);
    44       for (int i = 0; i < frontArr.Count; i++) {
    45         var d = Contribution(frontArr, i, problem.Maximization, refPoint, hv);
    46         if (min > d) {
    47           mindex = i;
    48           min = d;
    49         }
    50       }
    51 
    52       return mindex;
     35    public int LeastContributer<TR>(IEnumerable<TR> front, Func<TR, double[]> extractor, MultiObjectiveTestFunctionProblem problem) {
     36      var frontArr = front.Select(extractor.Invoke).ToList();
     37      var refPoint = BuildReference(frontArr.Concat(new[] { problem.TestFunction.ReferencePoint(problem.Objectives) }), problem.Maximization);
     38      var hv = Hypervolume.Calculate(frontArr, refPoint, problem.Maximization);
     39      return Enumerable.Range(0, frontArr.Count).Select(i => Contribution(frontArr, i, problem.Maximization, refPoint, hv)).ArgMin();
    5340    }
    5441
    55     private double Contribution(List<double[]> front, int idx, bool[] maximization, double[] refPoint, double hv) {
     42    private static double Contribution(IList<double[]> front, int idx, bool[] maximization, double[] refPoint, double hv) {
    5643      var point = front[idx];
    5744      front.RemoveAt(idx);
    … …  
    6047      return contribution;
    6148    }
    62 
    63     private double[] buildReference(IEnumerable<double[]> front, bool[] maximization) {
    64       double[] refPoint = new double[maximization.Length];
    65       foreach (double[] point in front) {
    66         for (int i = 0; i < maximization.Length; i++) {
     49    private static double[] BuildReference(IEnumerable<double[]> front, IReadOnlyList<bool> maximization) {
     50      var refPoint = new double[maximization.Count];
     51      foreach (var point in front)
     52        for (var i = 0; i < maximization.Count; i++)
    6753          refPoint[i] = maximization[i] ? Math.Min(refPoint[i], point[i]) : Math.Max(refPoint[i], point[i]);
    68         }
    69       }
    7054      return refPoint;
    71     }
    72     private double[] buildReference(double[] point1, double[] point2, bool[] maximization) {
    73       double[] refPoint = new double[maximization.Length];
    74 
    75       for (int i = 0; i < maximization.Length; i++) {
    76         refPoint[i] = maximization[i] ? Math.Min(point2[i], point1[i]) : Math.Max(point2[i], point1[i]);
    77       }
    78 
    79       return refPoint;
    80     }
    81 
    82     public override IDeepCloneable Clone(Cloner cloner) {
    83       return new HypervolumeIndicator(this, cloner);
    8455    }
    8556
    8657    [StorableConstructor]
    8758    protected HypervolumeIndicator(bool deserializing) : base(deserializing) { }
    88     protected HypervolumeIndicator(HypervolumeIndicator original, Cloner cloner)
    89       : base(original, cloner) {
    90     }
     59    protected HypervolumeIndicator(HypervolumeIndicator original, Cloner cloner) : base(original, cloner) { }
     60    public override IDeepCloneable Clone(Cloner cloner) { return new HypervolumeIndicator(this, cloner); }
    9161    public HypervolumeIndicator() { }
    9262  }

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/IIndicator.cs

    r14269 r14404  
    2020#endregion
    2121using System;
     22using System.Collections.Generic;
    2223using HeuristicLab.Core;
    2324using HeuristicLab.Problems.TestFunctions.MultiObjective;
    … …  
    2930    ///  A particularly good point may not be as benefical in another front.
    3031    /// </summary>
    31     /// <typeparam name="R"></typeparam>
     32    /// <typeparam name="TR"></typeparam>
    3233    /// <param name="front">a front which will be evaluated</param>
    3334    /// <param name="extractor">some mapping function to map any R to double[]</param>
    3435    /// <param name="problem">The problem on which the front is evaluated (!! The function itself will NOT be evluated only bounds referencePoints & other metadate will be used</param>
    3536    /// <returns>the index of the least contributing point according to any type of quality criteria</returns>
    36     int LeastContributer<R>(R[] front, Func<R, double[]> extractor, MultiObjectiveTestFunctionProblem problem);
     37    int LeastContributer<TR>(IEnumerable<TR> front, Func<TR, double[]> extractor, MultiObjectiveTestFunctionProblem problem);
    3738  }
    3839}

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/MOCMASEvolutionStrategy.cs

    r14269 r14404  
    2323using System;
    2424using System.Collections.Generic;
     25using System.Linq;
    2526using System.Threading;
     27using HeuristicLab.Algorithms.CMAEvolutionStrategy;
    2628using HeuristicLab.Analysis;
    2729using HeuristicLab.Common;
    … …  
    3234using HeuristicLab.Parameters;
    3335using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
     36using HeuristicLab.Problems.TestFunctions.MultiObjective;
    3437using HeuristicLab.Random;
    35 using System.Linq;
    36 using HeuristicLab.Problems.TestFunctions.MultiObjective;
    37 using HeuristicLab.Algorithms.MOCMAEvolutionStrategy;
    3838
    3939namespace HeuristicLab.Algorithms.MOCMAEvolutionStrategy {
    … …  
    4242  [StorableClass]
    4343  public class MOCMASEvolutionStrategy : BasicAlgorithm {
    44     public override Type ProblemType {
     44    public override Type ProblemType
     45    {
    4546      get { return typeof(MultiObjectiveTestFunctionProblem); }
    4647    }
    47     public new MultiObjectiveTestFunctionProblem Problem {
     48    public new MultiObjectiveTestFunctionProblem Problem
     49    {
    4850      get { return (MultiObjectiveTestFunctionProblem)base.Problem; }
    4951      set { base.Problem = value; }
    5052    }
     53    #region internal variables
     54    private readonly IRandom random = new MersenneTwister();
     55    private NormalDistributedRandom gauss;
     56    private MOCMAESIndividual[] solutions;
     57    private MOCMAESParameters internals;
     58    #endregion
    5159
    5260    #region ParameterNames
    … …  
    7684    #endregion
    7785
    78     #region internal variables
    79 
    80     private readonly IRandom random = new MersenneTwister();
    81     private NormalDistributedRandom gauss;
    82     private MOCMAESIndividual[] solutions;
    83     private const double penalizeFactor = 1e-6;
    84 
    85     //TODO OptionalValueParameter
    86     private double stepSizeLearningRate; //=cp learning rate in [0,1]
    87     private double stepSizeDampeningFactor; //d
    88     private double targetSuccessProbability;// p^target_succ
    89     private double evolutionPathLearningRate;//cc
    90     private double covarianceMatrixLearningRate;//ccov
    91     private double covarianceMatrixUnlearningRate;   //from shark
    92     private double successThreshold; //ptresh
    93     #endregion
    94 
    9586    #region ParameterProperties
    96     public IFixedValueParameter<IntValue> MaximumRuntimeParameter {
     87    public IFixedValueParameter<IntValue> MaximumRuntimeParameter
     88    {
    9789      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumRuntimeName]; }
    9890    }
    99     public IFixedValueParameter<IntValue> SeedParameter {
     91    public IFixedValueParameter<IntValue> SeedParameter
     92    {
    10093      get { return (IFixedValueParameter<IntValue>)Parameters[SeedName]; }
    10194    }
    102     public FixedValueParameter<BoolValue> SetSeedRandomlyParameter {
     95    public FixedValueParameter<BoolValue> SetSeedRandomlyParameter
     96    {
    10397      get { return (FixedValueParameter<BoolValue>)Parameters[SetSeedRandomlyName]; }
    10498    }
    105     private IFixedValueParameter<IntValue> PopulationSizeParameter {
     99    private IFixedValueParameter<IntValue> PopulationSizeParameter
     100    {
    106101      get { return (IFixedValueParameter<IntValue>)Parameters[PopulationSizeName]; }
    107102    }
    108     private IFixedValueParameter<IntValue> MaximumGenerationsParameter {
     103    private IFixedValueParameter<IntValue> MaximumGenerationsParameter
     104    {
    109105      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumGenerationsName]; }
    110106    }
    111     private IFixedValueParameter<IntValue> MaximumEvaluatedSolutionsParameter {
     107    private IFixedValueParameter<IntValue> MaximumEvaluatedSolutionsParameter
     108    {
    112109      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumEvaluatedSolutionsName]; }
    113110    }
    114     private IFixedValueParameter<DoubleValue> InitialSigmaParameter {
     111    private IFixedValueParameter<DoubleValue> InitialSigmaParameter
     112    {
    115113      get { return (IFixedValueParameter<DoubleValue>)Parameters[InitialSigmaName]; }
    116114    }
    117     private IConstrainedValueParameter<IIndicator> IndicatorParameter {
     115    private IConstrainedValueParameter<IIndicator> IndicatorParameter
     116    {
    118117      get { return (IConstrainedValueParameter<IIndicator>)Parameters[IndicatorName]; }
    119118    }
    … …  
    121120
    122121    #region Properties
    123     public int MaximumRuntime {
     122    public int MaximumRuntime
     123    {
    124124      get { return MaximumRuntimeParameter.Value.Value; }
    125125      set { MaximumRuntimeParameter.Value.Value = value; }
    126126    }
    127     public int Seed {
     127    public int Seed
     128    {
    128129      get { return SeedParameter.Value.Value; }
    129130      set { SeedParameter.Value.Value = value; }
    130131    }
    131     public bool SetSeedRandomly {
     132    public bool SetSeedRandomly
     133    {
    132134      get { return SetSeedRandomlyParameter.Value.Value; }
    133135      set { SetSeedRandomlyParameter.Value.Value = value; }
    134136    }
    135     public int PopulationSize {
     137    public int PopulationSize
     138    {
    136139      get { return PopulationSizeParameter.Value.Value; }
    137140      set { PopulationSizeParameter.Value.Value = value; }
    138141    }
    139     public int MaximumGenerations {
     142    public int MaximumGenerations
     143    {
    140144      get { return MaximumGenerationsParameter.Value.Value; }
    141145      set { MaximumGenerationsParameter.Value.Value = value; }
    142146    }
    143     public int MaximumEvaluatedSolutions {
     147    public int MaximumEvaluatedSolutions
     148    {
    144149      get { return MaximumEvaluatedSolutionsParameter.Value.Value; }
    145150      set { MaximumEvaluatedSolutionsParameter.Value.Value = value; }
    146151    }
    147     public double InitialSigma {
     152    public double InitialSigma
     153    {
    148154      get { return InitialSigmaParameter.Value.Value; }
    149155      set { InitialSigmaParameter.Value.Value = value; }
    150156    }
    151     public IIndicator Indicator {
     157    public IIndicator Indicator
     158    {
    152159      get { return IndicatorParameter.Value; }
    153160      set { IndicatorParameter.Value = value; }
    … …  
    156163
    157164    #endregion
     165
    158166    #region ResultsProperties
    159     private int ResultsEvaluations {
     167    private int ResultsEvaluations
     168    {
    160169      get { return ((IntValue)Results[EvaluationsResultName].Value).Value; }
    161170      set { ((IntValue)Results[EvaluationsResultName].Value).Value = value; }
    162171    }
    163     private int ResultsIterations {
     172    private int ResultsIterations
     173    {
    164174      get { return ((IntValue)Results[IterationsResultName].Value).Value; }
    165175      set { ((IntValue)Results[IterationsResultName].Value).Value = value; }
    166176    }
    167177    #region Datatable
    168     private DataTable ResultsQualities {
     178    private DataTable ResultsQualities
     179    {
    169180      get { return (DataTable)Results[TimetableResultName].Value; }
    170181    }
    171     private DataRow ResultsBestHypervolumeDataLine {
     182    private DataRow ResultsBestHypervolumeDataLine
     183    {
    172184      get { return ResultsQualities.Rows[BestHypervolumeResultName]; }
    173185    }
    174     private DataRow ResultsHypervolumeDataLine {
     186    private DataRow ResultsHypervolumeDataLine
     187    {
    175188      get { return ResultsQualities.Rows[HypervolumeResultName]; }
    176189    }
    177     private DataRow ResultsGenerationalDistanceDataLine {
     190    private DataRow ResultsGenerationalDistanceDataLine
     191    {
    178192      get { return ResultsQualities.Rows[GenerationalDistanceResultName]; }
    179193    }
    180     private DataRow ResultsInvertedGenerationalDistanceDataLine {
     194    private DataRow ResultsInvertedGenerationalDistanceDataLine
     195    {
    181196      get { return ResultsQualities.Rows[InvertedGenerationalDistanceResultName]; }
    182197    }
    183     private DataRow ResultsCrowdingDataLine {
     198    private DataRow ResultsCrowdingDataLine
     199    {
    184200      get { return ResultsQualities.Rows[CrowdingResultName]; }
    185201    }
    186     private DataRow ResultsSpacingDataLine {
     202    private DataRow ResultsSpacingDataLine
     203    {
    187204      get { return ResultsQualities.Rows[SpacingResultName]; }
    188205    }
    189     private DataRow ResultsHypervolumeDifferenceDataLine {
     206    private DataRow ResultsHypervolumeDifferenceDataLine
     207    {
    190208      get { return ResultsQualities.Rows[DifferenceToBestKnownHypervolumeResultName]; }
    191209    }
    192210    #endregion
    193211    //QualityIndicators
    194     private double ResultsHypervolume {
     212    private double ResultsHypervolume
     213    {
    195214      get { return ((DoubleValue)Results[HypervolumeResultName].Value).Value; }
    196215      set { ((DoubleValue)Results[HypervolumeResultName].Value).Value = value; }
    197216    }
    198     private double ResultsGenerationalDistance {
     217    private double ResultsGenerationalDistance
     218    {
    199219      get { return ((DoubleValue)Results[GenerationalDistanceResultName].Value).Value; }
    200220      set { ((DoubleValue)Results[GenerationalDistanceResultName].Value).Value = value; }
    201221    }
    202     private double ResultsInvertedGenerationalDistance {
     222    private double ResultsInvertedGenerationalDistance
     223    {
    203224      get { return ((DoubleValue)Results[InvertedGenerationalDistanceResultName].Value).Value; }
    204225      set { ((DoubleValue)Results[InvertedGenerationalDistanceResultName].Value).Value = value; }
    205226    }
    206     private double ResultsCrowding {
     227    private double ResultsCrowding
     228    {
    207229      get { return ((DoubleValue)Results[CrowdingResultName].Value).Value; }
    208230      set { ((DoubleValue)Results[CrowdingResultName].Value).Value = value; }
    209231    }
    210     private double ResultsSpacing {
     232    private double ResultsSpacing
     233    {
    211234      get { return ((DoubleValue)Results[SpacingResultName].Value).Value; }
    212235      set { ((DoubleValue)Results[SpacingResultName].Value).Value = value; }
    213236    }
    214     private double ResultsBestHypervolume {
     237    private double ResultsBestHypervolume
     238    {
    215239      get { return ((DoubleValue)Results[BestHypervolumeResultName].Value).Value; }
    216240      set { ((DoubleValue)Results[BestHypervolumeResultName].Value).Value = value; }
    217241    }
    218     private double ResultsBestKnownHypervolume {
     242    private double ResultsBestKnownHypervolume
     243    {
    219244      get { return ((DoubleValue)Results[BestKnownHypervolumeResultName].Value).Value; }
    220245      set { ((DoubleValue)Results[BestKnownHypervolumeResultName].Value).Value = value; }
    221246    }
    222     private double ResultsDifferenceBestKnownHypervolume {
     247    private double ResultsDifferenceBestKnownHypervolume
     248    {
    223249      get { return ((DoubleValue)Results[DifferenceToBestKnownHypervolumeResultName].Value).Value; }
    224250      set { ((DoubleValue)Results[DifferenceToBestKnownHypervolumeResultName].Value).Value = value; }
    … …  
    226252    }
    227253    //Solutions
    228     private DoubleMatrix ResultsSolutions {
     254    private DoubleMatrix ResultsSolutions
     255    {
    229256      get { return ((DoubleMatrix)Results[SolutionsResultName].Value); }
    230257      set { Results[SolutionsResultName].Value = value; }
    231258    }
    232     private ScatterPlotContent ResultsScatterPlot {
     259    private ScatterPlotContent ResultsScatterPlot
     260    {
    233261      get { return ((ScatterPlotContent)Results[ScatterPlotResultName].Value); }
    234262      set { Results[ScatterPlotResultName].Value = value; }
    … …  
    236264    #endregion
    237265
    238     #region constructors and hlBoilerPlate-code
    239     [StorableConstructor]
    240     protected MOCMASEvolutionStrategy(bool deserializing) : base(deserializing) { }
    241 
    242     protected MOCMASEvolutionStrategy(MOCMASEvolutionStrategy original, Cloner cloner)
    243         : base(original, cloner) {
    244     }
    245 
    246     public override IDeepCloneable Clone(Cloner cloner) {
    247       return new MOCMASEvolutionStrategy(this, cloner);
    248     }
    249 
     266    #region Constructors
    250267    public MOCMASEvolutionStrategy() {
    251268      Parameters.Add(new FixedValueParameter<IntValue>(MaximumRuntimeName, "The maximum runtime in seconds after which the algorithm stops. Use -1 to specify no limit for the runtime", new IntValue(3600)));
    … …  
    256273      Parameters.Add(new FixedValueParameter<IntValue>(MaximumGenerationsName, "The maximum number of generations which should be processed.", new IntValue(1000)));
    257274      Parameters.Add(new FixedValueParameter<IntValue>(MaximumEvaluatedSolutionsName, "The maximum number of evaluated solutions that should be computed.", new IntValue(int.MaxValue)));
    258 
    259       ItemSet<IIndicator> set = new ItemSet<IIndicator>();
    260       var default_ = new HypervolumeIndicator();
    261       set.Add(default_);
    262       set.Add(new CrowdingIndicator());
    263       Parameters.Add(new ConstrainedValueParameter<IIndicator>(IndicatorName, "The selection mechanism on non-dominated solutions", set, default_));
    264     }
    265     #endregion
    266 
    267     private void AddResult<T>(String name, String desc, T defaultValue) where T : class, IItem {
    268       Results.Add(new Result(name, desc, defaultValue));
    269     }
    270 
    271     #region updates
    272     private void UpdatePopulation(MOCMAESIndividual[] parents) {
    273       int[] offspringSucess = new int[solutions.Length];
    274       int offspringLength = parents.Length - solutions.Length;
    275 
    276       for (int i = 0; i < offspringLength; i++) {
    277         if (parents[i + solutions.Length].selected) {
    278           UpdateAsOffspring(parents[i + solutions.Length]);
    279 
    280           //TODO this may change if more offspring per parent is allowed
    281           offspringSucess[i] += MOCMAESIndividual.success;
    282         }
    283       }
    284       for (int i = 0; i < solutions.Length; i++) {
    285         if (parents[i].selected) {
    286           UpdateAsParent(parents[i], offspringSucess[i]);
    287         }
    288       }
    289 
    290       solutions = new MOCMAESIndividual[solutions.Length];
    291       int j = 0;
    292       foreach (MOCMAESIndividual ind in parents) {
    293         if (ind.selected) {
    294           solutions[j++] = ind;
    295         }
    296       }
    297 
    298     }
    299 
    300     private void UpdateAsParent(MOCMAESIndividual c, int v) {
    301       c.successProbability = (1 - stepSizeLearningRate) * c.successProbability + stepSizeLearningRate * (v == MOCMAESIndividual.success ? 1 : 0);
    302       c.sigma *= Math.Exp(1 / stepSizeDampeningFactor * (c.successProbability - targetSuccessProbability) / (1 - targetSuccessProbability));
    303       if (v != MOCMAESIndividual.failure) return;
    304       if (c.successProbability < successThreshold) {
    305         double stepNormSqr = c.GetSetpNormSqr();
    306         double rate = covarianceMatrixUnlearningRate;
    307         if (stepNormSqr > 1 && 1 < covarianceMatrixUnlearningRate * (2 * stepNormSqr - 1)) {
    308           rate = 1 / (2 * stepNormSqr - 1);
    309         }
    310         RankOneUpdate(c, 1 + rate, -rate, c.lastStep);
    311 
    312       } else {
    313         RoundUpdate(c);
    314       }
    315 
    316     }
    317 
    318     private void UpdateAsOffspring(MOCMAESIndividual c) {
    319       c.successProbability = (1 - stepSizeLearningRate) * c.successProbability + stepSizeLearningRate;
    320       c.sigma *= Math.Exp(1 / stepSizeDampeningFactor * (c.successProbability - targetSuccessProbability) / (1 - targetSuccessProbability));
    321       double evolutionpathUpdateWeight = evolutionPathLearningRate * (2.0 - evolutionPathLearningRate);
    322       if (c.successProbability < successThreshold) {
    323         c.UpdateEvolutionPath(1 - evolutionPathLearningRate, evolutionpathUpdateWeight);
    324         RankOneUpdate(c, 1 - covarianceMatrixLearningRate, covarianceMatrixLearningRate, c.evolutionPath);
    325       } else {
    326         RoundUpdate(c);
    327       }
    328     }
    329 
    330     private void RankOneUpdate(MOCMAESIndividual c, double v1, double v2, RealVector lastStep) {
    331       c.CholeskyUpdate(lastStep, v1, v2);
    332     }
    333 
    334     private void RoundUpdate(MOCMAESIndividual c) {
    335       double evolutionPathUpdateWeight = evolutionPathLearningRate * (2.0 - evolutionPathLearningRate);
    336       c.UpdateEvolutionPath(1 - evolutionPathLearningRate, 0);
    337       RankOneUpdate(c, 1 - covarianceMatrixLearningRate + evolutionPathUpdateWeight,
    338         covarianceMatrixLearningRate, c.evolutionPath);
    339     }
    340     #endregion
    341     #region selection
    342 
    343     private void Selection(MOCMAESIndividual[] parents, int length) {
    344       //perform a nondominated sort to assign the rank to every element
    345       var fronts = NonDominatedSort(parents);
    346 
    347       //deselect the highest rank fronts until we would end up with less or equal mu elements
    348       int rank = fronts.Count - 1;
    349       int popSize = parents.Length;
    350       while (popSize - fronts[rank].Count >= length) {
    351         var front = fronts[rank];
    352         foreach (var i in front) i.selected = false;
    353         popSize -= front.Count;
    354         rank--;
    355       }
    356 
    357       //now use the indicator to deselect the worst approximating elements of the last selected front
    358 
    359       var front_ = fronts[rank];
    360       for (; popSize > length; popSize--) {
    361         int lc = Indicator.LeastContributer<MOCMAESIndividual>(front_.ToArray(), x => x.penalizedFitness, Problem);
    362         front_[lc].selected = false;
    363         front_.Swap(lc, front_.Count - 1);
    364         front_.RemoveAt(front_.Count - 1);
    365       }
    366     }
    367     #endregion
    368     #region penalize Box-Constraints
    369     private void PenalizeEvaluate(IEnumerable<MOCMAESIndividual> offspring) {
    370       foreach (MOCMAESIndividual child in offspring) PenalizeEvaluate(child);
    371     }
    372 
    373     private void PenalizeEvaluate(MOCMAESIndividual offspring) {
    374       if (IsFeasable(offspring.x)) {
    375         offspring.fitness = Evaluate(offspring.x);
    376         offspring.penalizedFitness = offspring.fitness;
    377       } else {
    378         RealVector t = ClosestFeasible(offspring.x);
    379         offspring.fitness = Evaluate(t);
    380         offspring.penalizedFitness = Penalize(offspring.x, t, offspring.fitness);
    381       }
    382     }
    383 
    384     private double[] Evaluate(RealVector x) {
    385       double[] res = Problem.Evaluate(x);
    386       ResultsEvaluations++;
    387       return res;
    388     }
    389 
    390     private double[] Penalize(RealVector x, RealVector t, double[] fitness) {
    391       double penalty = Penalize(x, t);
    392       return fitness.Select(v => v + penalty).ToArray();
    393     }
    394 
    395     private double Penalize(RealVector x, RealVector t) {
    396       double sum = 0;
    397       for (int i = 0; i < x.Length; i++) {
    398         double d = x[i] - t[i];
    399         sum += d * d;
    400       }
    401       return sum;
    402     }
    403 
    404     private RealVector ClosestFeasible(RealVector x) {
    405       DoubleMatrix bounds = Problem.Bounds;
    406       RealVector r = new RealVector(x.Length);
    407 
    408       for (int i = 0; i < x.Length; i++) {
    409         int dim = i % bounds.Rows;
    410         r[i] = Math.Min(Math.Max(bounds[dim, 0], x[i]), bounds[dim, 1]);
    411       }
    412       return r;
    413     }
    414 
    415     private bool IsFeasable(RealVector offspring) {
    416       DoubleMatrix bounds = Problem.Bounds;
    417 
    418       for (int i = 0; i < offspring.Length; i++) {
    419         int dim = i % bounds.Rows;
    420         if (bounds[dim, 0] > offspring[i] || offspring[i] > bounds[dim, 1]) return false;
    421       }
    422       return true;
    423     }
    424     #endregion
    425 
    426     #region mutation
    427     private MOCMAESIndividual[] GenerateOffspring() {
    428       MOCMAESIndividual[] offspring = new MOCMAESIndividual[PopulationSize];
    429       for (int i = 0; i < PopulationSize; i++) {
    430         offspring[i] = new MOCMAESIndividual(solutions[i]);
    431         offspring[i].Mutate(gauss);
    432       }
    433       return offspring;
    434     }
    435     #endregion
    436 
    437     #region initialization
     275      var set = new ItemSet<IIndicator> { new HypervolumeIndicator(), new CrowdingIndicator() };
     276      Parameters.Add(new ConstrainedValueParameter<IIndicator>(IndicatorName, "The selection mechanism on non-dominated solutions", set, set.First()));
     277    }
     278
     279    [StorableConstructor]
     280    protected MOCMASEvolutionStrategy(bool deserializing) : base(deserializing) { }
     281    protected MOCMASEvolutionStrategy(MOCMASEvolutionStrategy original, Cloner cloner) : base(original, cloner) { }
     282    public override IDeepCloneable Clone(Cloner cloner) { return new MOCMASEvolutionStrategy(this, cloner); }
     283    #endregion
     284
     285    #region Mainloop
     286    protected override void Run(CancellationToken cancellationToken) {
     287      // Set up the algorithm
     288      if (SetSeedRandomly) Seed = new System.Random().Next();
     289      random.Reset(Seed);
     290      gauss = new NormalDistributedRandom(random, 0, 1);
     291
     292      InitResults();
     293      InitStrategy();
     294      InitSolutions();
     295      AnalyzeSolutions();
     296
     297      // Loop until iteration limit reached or canceled.
     298      for (ResultsIterations = 1; ResultsIterations < MaximumGenerations; ResultsIterations++) {
     299        try {
     300          Iterate();
     301          cancellationToken.ThrowIfCancellationRequested();
     302        }
     303        finally {
     304          AnalyzeSolutions();
     305        }
     306      }
     307    }
     308    private void Iterate() {
     309      var offspring = MutateOffspring();
     310      PenalizeEvaluate(offspring);
     311      var parents = solutions.Concat(offspring).ToArray();
     312      SelectParents(parents, solutions.Length);
     313      UpdatePopulation(parents);
     314    }
     315    protected override void OnExecutionTimeChanged() {
     316      base.OnExecutionTimeChanged();
     317      if (CancellationTokenSource == null) return;
     318      if (MaximumRuntime == -1) return;
     319      if (ExecutionTime.TotalSeconds > MaximumRuntime) CancellationTokenSource.Cancel();
     320    }
     321    #endregion
     322
     323    #region Initialization
    438324    private MOCMAESIndividual InitializeIndividual(RealVector x) {
    439325      var zeros = new RealVector(x.Length);
    440326      var identity = new double[x.Length, x.Length];
    441       for (int i = 0; i < x.Length; i++) {
    442         identity[i, i] = 1;
    443       }
    444       return new MOCMAESIndividual(x, targetSuccessProbability, InitialSigma, zeros, identity);
    445     }
    446 
     327      for (var i = 0; i < x.Length; i++) identity[i, i] = 1;
     328      return new MOCMAESIndividual(x, internals.TargetSuccessProbability, InitialSigma, zeros, identity, internals);
     329    }
    447330    private void InitSolutions() {
    448331      solutions = new MOCMAESIndividual[PopulationSize];
    449       for (int i = 0; i < PopulationSize; i++) {
    450         RealVector x = new RealVector(Problem.ProblemSize); // Uniform distibution in all dimesions assumed.
    451                                                             // There is the UniformSolutionCreater associated with the Encoding, but it was considered not usable here
     332      for (var i = 0; i < PopulationSize; i++) {
     333        var x = new RealVector(Problem.ProblemSize); // Uniform distibution in all dimensions assumed.
     334                                                     // There is the UniformSolutionCreater associated with the Encoding, but it was considered not usable here
    452335        var bounds = Problem.Bounds;
    453         for (int j = 0; j < Problem.Objectives; j++) {
    454           int dim = j % bounds.Rows;
     336        for (var j = 0; j < Problem.ProblemSize; j++) {
     337          var dim = j % bounds.Rows;
    455338          x[j] = random.NextDouble() * (bounds[dim, 1] - bounds[dim, 0]) + bounds[dim, 0];
    456339        }
    … …  
    459342      PenalizeEvaluate(solutions);
    460343    }
    461 
    462344    private void InitStrategy() {
    463       int lambda = 1;
     345      const int lambda = 1;
    464346      double n = Problem.ProblemSize;
    465       gauss = new NormalDistributedRandom(random, 0, 1);
    466       targetSuccessProbability = 1.0 / (5.0 + Math.Sqrt(lambda) / 2.0);
    467       stepSizeDampeningFactor = 1.0 + n / (2.0 * lambda);
    468       stepSizeLearningRate = targetSuccessProbability * lambda / (2.0 + targetSuccessProbability * lambda);
    469       evolutionPathLearningRate = 2.0 / (n + 2.0);
    470       covarianceMatrixLearningRate = 2.0 / (n * n + 6.0);
    471       covarianceMatrixUnlearningRate = 0.4 / (Math.Pow(n, 1.6) + 1);
    472       successThreshold = 0.44;
    473 
    474     }
    475 
     347
     348      var targetSuccessProbability = 1.0 / (5.0 + Math.Sqrt(lambda) / 2.0);
     349      var stepSizeDampeningFactor = 1.0 + n / (2.0 * lambda);
     350      var stepSizeLearningRate = targetSuccessProbability * lambda / (2.0 + targetSuccessProbability * lambda);
     351      var evolutionPathLearningRate = 2.0 / (n + 2.0);
     352      var covarianceMatrixLearningRate = 2.0 / (n * n + 6.0);
     353      var covarianceMatrixUnlearningRate = 0.4 / (Math.Pow(n, 1.6) + 1);
     354      var successThreshold = 0.44;
     355      internals = new MOCMAESParameters(stepSizeLearningRate, stepSizeDampeningFactor, targetSuccessProbability, evolutionPathLearningRate, covarianceMatrixLearningRate, covarianceMatrixUnlearningRate, successThreshold);
     356
     357    }
    476358    private void InitResults() {
    477359      AddResult(IterationsResultName, "The number of gererations evaluated", new IntValue(0));
    … …  
    479361      AddResult(HypervolumeResultName, "The hypervolume of the current front considering the Referencepoint defined in the Problem", new DoubleValue(0.0));
    480362      AddResult(BestHypervolumeResultName, "The best hypervolume of the current run considering the Referencepoint defined in the Problem", new DoubleValue(0.0));
    481       AddResult(BestKnownHypervolumeResultName, "The best knwon hypervolume considering the Referencepoint defined in the Problem", new DoubleValue(Double.NaN));
    482       AddResult(DifferenceToBestKnownHypervolumeResultName, "The difference between the current and the best known hypervolume", new DoubleValue(Double.NaN));
    483       AddResult(GenerationalDistanceResultName, "The generational distance to an optimal pareto front defined in the Problem", new DoubleValue(Double.NaN));
    484       AddResult(InvertedGenerationalDistanceResultName, "The inverted generational distance to an optimal pareto front defined in the Problem", new DoubleValue(Double.NaN));
     363      AddResult(BestKnownHypervolumeResultName, "The best knwon hypervolume considering the Referencepoint defined in the Problem", new DoubleValue(double.NaN));
     364      AddResult(DifferenceToBestKnownHypervolumeResultName, "The difference between the current and the best known hypervolume", new DoubleValue(double.NaN));
     365      AddResult(GenerationalDistanceResultName, "The generational distance to an optimal pareto front defined in the Problem", new DoubleValue(double.NaN));
     366      AddResult(InvertedGenerationalDistanceResultName, "The inverted generational distance to an optimal pareto front defined in the Problem", new DoubleValue(double.NaN));
    485367      AddResult(CrowdingResultName, "The average crowding value for the current front (excluding infinities)", new DoubleValue(0.0));
    486368      AddResult(SpacingResultName, "The spacing for the current front (excluding infinities)", new DoubleValue(0.0));
    … …  
    506388    #endregion
    507389
    508     #region analyze
     390    private MOCMAESIndividual[] MutateOffspring() {
     391      var offspring = new MOCMAESIndividual[PopulationSize];
     392      for (var i = 0; i < PopulationSize; i++) {
     393        offspring[i] = new MOCMAESIndividual(solutions[i]);
     394        offspring[i].Mutate(gauss);
     395      }
     396      return offspring;
     397    }
     398
     399    #region Evaluation
     400    private void PenalizeEvaluate(IEnumerable<MOCMAESIndividual> offspring) {
     401      foreach (var child in offspring) {
     402        if (IsFeasable(child.Mean)) {
     403          child.Fitness = Evaluate(child.Mean);
     404          child.PenalizedFitness = child.Fitness;
     405        } else {
     406          var t = ClosestFeasible(child.Mean);
     407          child.Fitness = Evaluate(t);
     408          child.PenalizedFitness = Penalize(child.Mean, t, child.Fitness);
     409        }
     410      }
     411    }
     412    private double[] Evaluate(RealVector x) {
     413      var res = Problem.Evaluate(x);
     414      ResultsEvaluations++;
     415      return res;
     416    }
     417    private static double[] Penalize(RealVector x, RealVector t, IEnumerable<double> fitness) {
     418      var penalty = x.Select((t1, i) => t1 - t[i]).Sum(d => d * d);
     419      return fitness.Select(v => v + penalty).ToArray();
     420    }
     421    private RealVector ClosestFeasible(RealVector x) {
     422      var bounds = Problem.Bounds;
     423      var r = new RealVector(x.Length);
     424      for (var i = 0; i < x.Length; i++) {
     425        var dim = i % bounds.Rows;
     426        r[i] = Math.Min(Math.Max(bounds[dim, 0], x[i]), bounds[dim, 1]);
     427      }
     428      return r;
     429    }
     430    private bool IsFeasable(RealVector offspring) {
     431      var bounds = Problem.Bounds;
     432      for (var i = 0; i < offspring.Length; i++) {
     433        var dim = i % bounds.Rows;
     434        if (bounds[dim, 0] > offspring[i] || offspring[i] > bounds[dim, 1]) return false;
     435      }
     436      return true;
     437    }
     438    #endregion
     439
     440    private void SelectParents(IReadOnlyList<MOCMAESIndividual> parents, int length) {
     441      //perform a nondominated sort to assign the rank to every element
     442      var fronts = NonDominatedSort(parents);
     443
     444      //deselect the highest rank fronts until we would end up with less or equal mu elements
     445      var rank = fronts.Count - 1;
     446      var popSize = parents.Count;
     447      while (popSize - fronts[rank].Count >= length) {
     448        var front = fronts[rank];
     449        foreach (var i in front) i.Selected = false;
     450        popSize -= front.Count;
     451        rank--;
     452      }
     453
     454      //now use the indicator to deselect the approximatingly worst elements of the last selected front
     455      var front1 = fronts[rank];
     456      for (; popSize > length; popSize--) {
     457        var lc = Indicator.LeastContributer(front1.ToArray(), x => x.PenalizedFitness, Problem);
     458        front1[lc].Selected = false;
     459        front1.Swap(lc, front1.Count - 1);
     460        front1.RemoveAt(front1.Count - 1);
     461      }
     462    }
     463
     464    private void UpdatePopulation(IReadOnlyList<MOCMAESIndividual> parents) {
     465      var offspringSucess = new int[solutions.Length];
     466      var offspringLength = parents.Count - solutions.Length;
     467      for (var i = 0; i < offspringLength; i++) {
     468        if (!parents[i + solutions.Length].Selected) continue;
     469        parents[i + solutions.Length].UpdateAsOffspring();
     470        offspringSucess[i] += MOCMAESIndividual.Success;
     471      }
     472      for (var i = 0; i < solutions.Length; i++) if (parents[i].Selected) parents[i].UpdateAsParent(offspringSucess[i]);
     473      solutions = new MOCMAESIndividual[solutions.Length];
     474      var j = 0;
     475      foreach (var ind in parents) if (ind.Selected) solutions[j++] = ind;
     476    }
     477
     478    #region Analysis
    509479    private void AnalyzeSolutions() {
    510       //solutions
    511       ResultsScatterPlot = new ScatterPlotContent(solutions.Select(x => x.fitness).ToArray<double[]>(),
    512           solutions.Select(x => ToArray(x.x)).ToArray<double[]>(),
     480      ResultsScatterPlot = new ScatterPlotContent(solutions.Select(x => x.Fitness).ToArray(),
     481          solutions.Select(x => x.Mean.ToArray()).ToArray(),
    513482          ResultsScatterPlot.ParetoFront,
    514483          ResultsScatterPlot.Objectives);
    515       ResultsSolutions = ToMatrix(solutions.Select(x => ToArray(x.x)));
     484      ResultsSolutions = ToMatrix(solutions.Select(x => x.Mean.ToArray()));
    516485      AnalyzeQualityIndicators();
    517 
    518     }
    519 
     486    }
    520487    private void AnalyzeQualityIndicators() {
    521 
    522       //var front = NonDominatedSelect.selectNonDominated(solutions.Select(x => x.fitness), Problem.Maximization, true);
    523       var front = NonDominatedSelect.GetDominatingVectors(solutions.Select(x => x.fitness), Problem.TestFunction.ReferencePoint(Problem.Objectives), Problem.Maximization, true);
    524       //front = NonDominatedSelect.removeNonReferenceDominatingVectors(front, Problem.TestFunction.ReferencePoint(Problem.Objectives), Problem.Maximization, false);
    525       var bounds = ToArray(Problem.Bounds);
     488      var front = NonDominatedSelect.GetDominatingVectors(solutions.Select(x => x.Fitness), Problem.TestFunction.ReferencePoint(Problem.Objectives), Problem.Maximization, true).ToArray();
     489      var bounds = Problem.Bounds.CloneAsMatrix();
    526490      ResultsCrowding = Crowding.Calculate(front, bounds);
    527491      ResultsSpacing = Spacing.Calculate(front);
    528       ResultsGenerationalDistance = Problem.BestKnownFront != null ? GenerationalDistance.Calculate(front, Utilities.ToArray(Problem.BestKnownFront), 1) : Double.NaN;
    529 
    530       ResultsInvertedGenerationalDistance = Problem.BestKnownFront != null ? InvertedGenerationalDistance.Calculate(front, Utilities.ToArray(Problem.BestKnownFront), 1) : Double.NaN;
     492      ResultsGenerationalDistance = Problem.BestKnownFront != null ? GenerationalDistance.Calculate(front, Utilities.ToArray(Problem.BestKnownFront), 1) : double.NaN;
     493
     494      ResultsInvertedGenerationalDistance = Problem.BestKnownFront != null ? InvertedGenerationalDistance.Calculate(front, Utilities.ToArray(Problem.BestKnownFront), 1) : double.NaN;
    531495
    532496      ResultsHypervolume = Hypervolume.Calculate(front, Problem.TestFunction.ReferencePoint(Problem.Objectives), Problem.Maximization);
    533497      ResultsBestHypervolume = Math.Max(ResultsHypervolume, ResultsBestHypervolume);
    534       ResultsDifferenceBestKnownHypervolume = ResultsBestKnownHypervolume - ResultsBestHypervolume; //Take best of this run or current HV???
     498      ResultsDifferenceBestKnownHypervolume = ResultsBestKnownHypervolume - ResultsBestHypervolume;
    535499
    536500      //Datalines
    … …  
    546510    #endregion
    547511
    548     //MU = populationSize
    549     #region mainloop
    550     protected override void Run(CancellationToken cancellationToken) {
    551       // Set up the algorithm
    552       if (SetSeedRandomly) Seed = new System.Random().Next();
    553       random.Reset(Seed);
    554 
    555       InitResults();
    556       InitStrategy();
    557       InitSolutions();
    558 
    559       // Loop until iteration limit reached or canceled.
    560       for (ResultsIterations = 1; ResultsIterations < MaximumGenerations; ResultsIterations++) {
    561         try {
    562           Iterate();
    563           cancellationToken.ThrowIfCancellationRequested();
    564         }
    565         finally {
    566           AnalyzeSolutions();
    567         }
    568       }
    569     }
    570 
    571     protected override void OnExecutionTimeChanged() {
    572       base.OnExecutionTimeChanged();
    573       if (CancellationTokenSource == null) return;
    574       if (MaximumRuntime == -1) return;
    575       if (ExecutionTime.TotalSeconds > MaximumRuntime) CancellationTokenSource.Cancel();
    576     }
    577 
    578     private void Iterate() {
    579       MOCMAESIndividual[] offspring = GenerateOffspring();
    580       PenalizeEvaluate(offspring);
    581       var parents = Merge(solutions, offspring);
    582       Selection(parents, solutions.Length);
    583       UpdatePopulation(parents);
    584 
    585     }
    586     #endregion
    587 
    588     private class MOCMAESIndividual {
    589       public static readonly int success = 1;
    590       public static readonly int noSuccess = 2;
    591       public static readonly int failure = 3;
    592 
    593 
    594       //Chromosome
    595       public RealVector x;
    596       public double successProbability;
    597       public double sigma;//stepsize
    598       public RealVector evolutionPath; // pc
    599       public RealVector lastStep;
    600       public RealVector lastZ;
    601       private double[,] lowerCholesky;
    602 
    603 
    604       //Phenotype
    605       public double[] fitness;
    606       public double[] penalizedFitness;
    607       public bool selected = true;
    608 
    609       internal double rank;
    610 
    611       /// <summary>
    612       ///
    613       /// </summary>
    614       /// <param name="x">has to be 0-vector with correct lenght</param>
    615       /// <param name="p_succ">has to be ptargetsucc</param>
    616       /// <param name="sigma">initialSigma</param>
    617       /// <param name="pc">has to be 0-vector with correct lenght</param>
    618       /// <param name="C">has to be a symmetric positive definit Covariance matrix</param>
    619       public MOCMAESIndividual(RealVector x, double p_succ, double sigma, RealVector pc, double[,] C) {
    620         this.x = x;
    621         this.successProbability = p_succ;
    622         this.sigma = sigma;
    623         this.evolutionPath = pc;
    624         CholeskyDecomposition(C);
    625       }
    626 
    627       private void CholeskyDecomposition(double[,] C) {
    628         if (!alglib.spdmatrixcholesky(ref C, C.GetLength(0), false)) {
    629           throw new ArgumentException("Covariancematrix is not symmetric positiv definit");
    630         }
    631         lowerCholesky = (double[,])C.Clone();
    632       }
    633 
    634       public MOCMAESIndividual(MOCMAESIndividual other) {
    635         this.successProbability = other.successProbability;
    636         this.sigma = other.sigma;
    637 
    638         this.evolutionPath = (RealVector)other.evolutionPath.Clone();
    639         this.x = (RealVector)other.x.Clone();
    640 
    641         this.lowerCholesky = (double[,])other.lowerCholesky.Clone();
    642       }
    643 
    644       public void UpdateEvolutionPath(double learningRate, double updateWeight) {
    645         updateWeight = Math.Sqrt(updateWeight);
    646         for (int i = 0; i < evolutionPath.Length; i++) {
    647           evolutionPath[i] *= learningRate;
    648           evolutionPath[i] += updateWeight * lastStep[i];
    649         }
    650       }
    651 
    652       public double GetSetpNormSqr() {
    653         double sum = 0;
    654         foreach (double d in lastZ) {
    655           sum += d * d;
    656         }
    657         return sum;
    658       }
    659 
    660       public void CholeskyUpdate(RealVector v, double alpha, double beta) {
    661         int n = v.Length;
    662         double[] temp = new double[n];
    663         for (int i = 0; i < n; i++) temp[i] = v[i];
    664         double betaPrime = 1;
    665         double a = Math.Sqrt(alpha);
    666         for (int j = 0; j < n; j++) {
    667           double Ljj = a * lowerCholesky[j, j];
    668           double dj = Ljj * Ljj;
    669           double wj = temp[j];
    670           double swj2 = beta * wj * wj;
    671           double gamma = dj * betaPrime + swj2;
    672           double x = dj + swj2 / betaPrime;
    673           if (x < 0.0) throw new ArgumentException("Update makes Covariancematrix indefinite");
    674           double nLjj = Math.Sqrt(x);
    675           lowerCholesky[j, j] = nLjj;
    676           betaPrime += swj2 / dj;
    677           if (j + 1 < n) {
    678             for (int i = j + 1; i < n; i++) {
    679               lowerCholesky[i, j] *= a;
    680             }
    681             for (int i = j + 1; i < n; i++) {
    682               temp[i] = wj / Ljj * lowerCholesky[i, j];
    683             }
    684             if (gamma == 0) continue;
    685             for (int i = j + 1; i < n; i++) {
    686               lowerCholesky[i, j] *= nLjj / Ljj;
    687             }
    688             for (int i = j + 1; i < n; i++) {
    689               lowerCholesky[i, j] += (nLjj * beta * wj / gamma) * temp[i];
    690             }
    691           }
    692 
    693         }
    694 
    695       }
    696 
    697       public void Mutate(NormalDistributedRandom gauss) {
    698 
    699         //sampling a random z from N(0,I) where I is the Identity matrix;
    700         lastZ = new RealVector(x.Length);
    701         int n = lastZ.Length;
    702         for (int i = 0; i < n; i++) {
    703           lastZ[i] = gauss.NextDouble();
    704         }
    705         //Matrixmultiplication: lastStep = lowerCholesky * lastZ;
    706         lastStep = new RealVector(x.Length);
    707         for (int i = 0; i < n; i++) {
    708           double sum = 0;
    709           for (int j = 0; j <= i; j++) {
    710             sum += lowerCholesky[i, j] * lastZ[j];
    711           }
    712           lastStep[i] = sum;
    713         }
    714 
    715         //add the step to x weighted by stepsize;
    716         for (int i = 0; i < x.Length; i++) {
    717           x[i] += sigma * lastStep[i];
    718         }
    719 
    720       }
    721 
    722     }
    723 
     512
     513
     514    #region Helpers
     515    public DoubleMatrix ToMatrix(IEnumerable<double[]> data) {
     516      var d2 = data.ToArray();
     517      var mat = new DoubleMatrix(d2.Length, d2[0].Length);
     518      for (var i = 0; i < mat.Rows; i++) {
     519        for (var j = 0; j < mat.Columns; j++) {
     520          mat[i, j] = d2[i][j];
     521        }
     522      }
     523      return mat;
     524    }
     525    private void AddResult<T>(string name1, string desc, T defaultValue) where T : class, IItem {
     526      Results.Add(new Result(name1, desc, defaultValue));
     527    }
    724528    //blatantly stolen form HeuristicLab.Optimization.Operators.FastNonDominatedSort
     529    //however: Operators.FastNonDominatedSort does not return ranked fronts => rerank after sorting would not be sensible
    725530    #region FastNonDominatedSort   
    726531    private enum DominationResult { Dominates, IsDominated, IsNonDominated };
    727 
    728     private List<List<MOCMAESIndividual>> NonDominatedSort(MOCMAESIndividual[] individuals) {
    729       bool dominateOnEqualQualities = false;
    730       bool[] maximization = Problem.Maximization;
     532    private List<List<MOCMAESIndividual>> NonDominatedSort(IReadOnlyList<MOCMAESIndividual> individuals) {
     533      const bool dominateOnEqualQualities = false;
     534      var maximization = Problem.Maximization;
    731535      if (individuals == null) throw new InvalidOperationException(Name + ": No qualities found.");
    732       int populationSize = individuals.Length;
    733 
    734       List<List<MOCMAESIndividual>> fronts = new List<List<MOCMAESIndividual>>();
    735       Dictionary<MOCMAESIndividual, List<int>> dominatedScopes = new Dictionary<MOCMAESIndividual, List<int>>();
    736       int[] dominationCounter = new int[populationSize];
    737 
    738       for (int pI = 0; pI < populationSize - 1; pI++) {
    739         MOCMAESIndividual p = individuals[pI];
     536      var populationSize = individuals.Count;
     537
     538      var fronts = new List<List<MOCMAESIndividual>>();
     539      var dominatedScopes = new Dictionary<MOCMAESIndividual, List<int>>();
     540      var dominationCounter = new int[populationSize];
     541
     542      for (var pI = 0; pI < populationSize - 1; pI++) {
     543        var p = individuals[pI];
    740544        List<int> dominatedScopesByp;
    741545        if (!dominatedScopes.TryGetValue(p, out dominatedScopesByp))
    742546          dominatedScopes[p] = dominatedScopesByp = new List<int>();
    743         for (int qI = pI + 1; qI < populationSize; qI++) {
    744           DominationResult test = Dominates(individuals[pI], individuals[qI], maximization, dominateOnEqualQualities);
     547        for (var qI = pI + 1; qI < populationSize; qI++) {
     548          var test = Dominates(individuals[pI], individuals[qI], maximization, dominateOnEqualQualities);
    745549          if (test == DominationResult.Dominates) {
    746550            dominatedScopesByp.Add(qI);
    … …  
    762566        }
    763567      }
    764       int i = 0;
     568      var i = 0;
    765569      while (i < fronts.Count && fronts[i].Count > 0) {
    766         List<MOCMAESIndividual> nextFront = new List<MOCMAESIndividual>();
    767         foreach (MOCMAESIndividual p in fronts[i]) {
     570        var nextFront = new List<MOCMAESIndividual>();
     571        foreach (var p in fronts[i]) {
    768572          List<int> dominatedScopesByp;
    769           if (dominatedScopes.TryGetValue(p, out dominatedScopesByp)) {
    770             for (int k = 0; k < dominatedScopesByp.Count; k++) {
    771               int dominatedScope = dominatedScopesByp[k];
    772               dominationCounter[dominatedScope] -= 1;
    773               if (dominationCounter[dominatedScope] == 0) {
    774                 nextFront.Add(individuals[dominatedScope]);
    775               }
    776             }
     573          if (!dominatedScopes.TryGetValue(p, out dominatedScopesByp)) continue;
     574          foreach (var dominatedScope in dominatedScopesByp) {
     575            dominationCounter[dominatedScope] -= 1;
     576            if (dominationCounter[dominatedScope] != 0) continue;
     577            nextFront.Add(individuals[dominatedScope]);
    777578          }
    778579        }
    … …  
    781582      }
    782583
    783       MOCMAESIndividual[] result = new MOCMAESIndividual[individuals.Length];
    784 
    785584      for (i = 0; i < fronts.Count; i++) {
    786585        foreach (var p in fronts[i]) {
    787           p.rank = i;
     586          p.Rank = i;
    788587        }
    789588      }
    790589      return fronts;
    791590    }
    792 
    793     private static void AddToFront(MOCMAESIndividual p, List<List<MOCMAESIndividual>> fronts, int i) {
     591    private static void AddToFront(MOCMAESIndividual p, IList<List<MOCMAESIndividual>> fronts, int i) {
    794592      if (i == fronts.Count) fronts.Add(new List<MOCMAESIndividual>());
    795593      fronts[i].Add(p);
    796594    }
    797 
    798595    private static DominationResult Dominates(MOCMAESIndividual left, MOCMAESIndividual right, bool[] maximizations, bool dominateOnEqualQualities) {
    799       return Dominates(left.penalizedFitness, right.penalizedFitness, maximizations, dominateOnEqualQualities);
    800     }
    801 
    802     private static DominationResult Dominates(double[] left, double[] right, bool[] maximizations, bool dominateOnEqualQualities) {
     596      return Dominates(left.PenalizedFitness, right.PenalizedFitness, maximizations, dominateOnEqualQualities);
     597    }
     598    private static DominationResult Dominates(IReadOnlyList<double> left, IReadOnlyList<double> right, IReadOnlyList<bool> maximizations, bool dominateOnEqualQualities) {
    803599      //mkommend Caution: do not use LINQ.SequenceEqual for comparing the two quality arrays (left and right) due to performance reasons
    804600      if (dominateOnEqualQualities) {
    805601        var equal = true;
    806         for (int i = 0; i < left.Length; i++) {
     602        for (var i = 0; i < left.Count; i++) {
    807603          if (left[i] != right[i]) {
    808604            equal = false;
    … …  
    814610
    815611      bool leftIsBetter = false, rightIsBetter = false;
    816       for (int i = 0; i < left.Length; i++) {
     612      for (var i = 0; i < left.Count; i++) {
    817613        if (IsDominated(left[i], right[i], maximizations[i])) rightIsBetter = true;
    818614        else if (IsDominated(right[i], left[i], maximizations[i])) leftIsBetter = true;
    … …  
    824620      return DominationResult.IsNonDominated;
    825621    }
    826 
    827622    private static bool IsDominated(double left, double right, bool maximization) {
    828623      return maximization && left < right
    829624        || !maximization && left > right;
    830625    }
    831 
    832     #endregion
    833 
    834     #region conversions
    835 
    836     private T[] Merge<T>(T[] parents, T[] offspring) {
    837       T[] merged = new T[parents.Length + offspring.Length];
    838       for (int i = 0; i < parents.Length; i++) {
    839         merged[i] = parents[i];
    840       }
    841       for (int i = 0; i < offspring.Length; i++) {
    842         merged[i + parents.Length] = offspring[i];
    843       }
    844       return merged;
    845     }
    846 
    847     public double[] ToArray(RealVector r) {
    848       double[] d = new double[r.Length];
    849       for (int i = 0; i < r.Length; i++) {
    850         d[i] = r[i];
    851       }
    852       return d;
    853     }
    854     public DoubleMatrix ToMatrix(IEnumerable<double[]> data) {
    855       var d2 = data.ToArray<double[]>();
    856       DoubleMatrix mat = new DoubleMatrix(d2.Length, d2[0].Length);
    857       for (int i = 0; i < mat.Rows; i++) {
    858         for (int j = 0; j < mat.Columns; j++) {
    859           mat[i, j] = d2[i][j];
    860         }
    861       }
    862       return mat;
    863     }
    864     public double[,] ToArray(DoubleMatrix data) {
    865 
    866       double[,] mat = new double[data.Rows, data.Columns];
    867       for (int i = 0; i < data.Rows; i++) {
    868         for (int j = 0; j < data.Columns; j++) {
    869           mat[i, j] = data[i, j];
    870         }
    871       }
    872       return mat;
    873     }
     626    #endregion
    874627    #endregion
    875628  }

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/Plugin.cs.frame

    r14269 r14404  
    2626  /// Plugin class for HeuristicLab.Algorithms.EvolutionStrategy plugin.
    2727  /// </summary>
    28   [Plugin("HeuristicLab.Algorithms.MOCMAEvolutionStrategy", "3.3.23.$WCREV$")]
     28  [Plugin("HeuristicLab.Algorithms.MOCMAEvolutionStrategy", "3.3.14.$WCREV$")]
    2929  [PluginFile("HeuristicLab.Algorithms.MOCMAEvolutionStrategy-3.3.dll", PluginFileType.Assembly)]
    3030  [PluginDependency("HeuristicLab.ALGLIB", "3.7")]
    … …  
    4242  [PluginDependency("HeuristicLab.Persistence", "3.3")]
    4343  [PluginDependency("HeuristicLab.Problems.Instances", "3.3")]
    44   [PluginDependency("HeuristicLab.Problems.MultiObjectiveTestFunctions", "3.3")]
     44  [PluginDependency("HeuristicLab.Problems.TestFunctions.MultiObjective", "3.3")]
    4545  [PluginDependency("HeuristicLab.Random", "3.3")]
    4646  public class HeuristicLabAlgorithmsMOCMAEvolutionStrategyPlugin : PluginBase {

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/Properties/AssemblyInfo.cs.frame

    r14269 r14404  
    5353// by using the '*' as shown below:
    5454[assembly: AssemblyVersion("3.3.0.0")]
    55 [assembly: AssemblyFileVersion("3.3.23.$WCREV$")]
     55[assembly: AssemblyFileVersion("3.3.14.$WCREV$")]

  • branches/MOCMAEvolutionStrategy/HeuristicLab.Algorithms.MOCMAEvolutionStrategy/3.3/Utilities.cs

    r14269 r14404  
    1 using System;
    2 using System.Collections.Generic;
    3 using System.Linq;
    4 using System.Text;
    5 using System.Threading.Tasks;
     1using System.Collections.Generic;
    62using HeuristicLab.Data;
    73
    84namespace HeuristicLab.Algorithms.MOCMAEvolutionStrategy {
    9   internal class Utilities {
     5  internal static class Utilities {
    106    internal static double[][] ToArray(DoubleMatrix m) {
    11       int i = m.Rows - 1;
    12       double[][] a = new double[i][];
     7      var i = m.Rows - 1;
     8      var a = new double[i][];
    139      for (i--; i >= 0; i--) {
    14         int j = m.Columns;
     10        var j = m.Columns;
    1511        a[i] = new double[j];
    1612        for (j--; j >= 0; j--) a[i][j] = m[i, j];
    … …  
    1915    }
    2016
     17    public static int ArgMin<T>(IEnumerable<T> values, IComparer<T> comparer) {
     18      var mindex = 0;
     19      var min = default(T);
     20      var i = 0;
     21      foreach (var v in values) {
     22        if (mindex < 0 || comparer.Compare(v, min) < 0) {
     23          min = v;
     24          mindex = i;
     25        }
     26        i++;
     27      }
     28      return mindex;
     29
     30    }
     31    public static int ArgMax<T>(IEnumerable<T> values, IComparer<T> comparer) {
     32      var mindex = 0;
     33      var min = default(T);
     34      var i = 0;
     35      foreach (var v in values) {
     36        if (mindex < 0 || comparer.Compare(v, min) > 0) {
     37          min = v;
     38          mindex = i;
     39        }
     40        i++;
     41      }
     42      return mindex;
     43    }
     44    private class DoubleComparer : IComparer<double> {
     45      public int Compare(double x, double y) {
     46        return x.CompareTo(y);
     47      }
     48    }
     49    public static int ArgMax(this IEnumerable<double> values) {
     50      return ArgMax(values, new DoubleComparer());
     51    }
     52    public static int ArgMin(this IEnumerable<double> values) {
     53      return ArgMin(values, new DoubleComparer());
     54    }
    2155  }
    2256}
Note: See TracChangeset for help on using the changeset viewer.