source: branches/2988_ModelsOfModels2/HeuristicLab.Algorithms.EMM/SelfConfiguration.cs @ 17625

using HEAL.Attic;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using HeuristicLab.Problems.DataAnalysis.Symbolic;
using System.Collections.Generic;
using System.Linq;

namespace HeuristicLab.Algorithms.EvolvmentModelsOfModels {
  [Item("SelfConfiguration", "A fluent realization of self-configuration mechanism")]
  [StorableType("7888A6B9-CDC8-42A0-BDA6-F9D89A56067A")]
  public class SelfConfiguration : Item {
    #region data members
    [Storable]
    public List<double> Probabilities { get; private set; }
    [Storable]
    protected List<List<double>> SucsessStatistics { get; private set; }
    [Storable]
    public double SocialKarte { get; private set; } // parameters of method
    [Storable]
    public double StepValue { get; private set; }
    [Storable]
    protected List<int> ChoiseForPopulation { get; private set; }
    #endregion
    #region constructors
    public SelfConfiguration() : base() {
      SucsessStatistics = new List<List<double>>();
      Probabilities = new List<double>();
      ChoiseForPopulation = new List<int>();
    }
    public SelfConfiguration(SelfConfiguration original, Cloner cloner) : base(original, cloner) {
      SucsessStatistics = original.SucsessStatistics.Select(x => x.ToList()).ToList();
      Probabilities = original.Probabilities.ToList();
      SocialKarte = original.SocialKarte;
      StepValue = original.StepValue;
    }
    [StorableConstructor]
    protected SelfConfiguration(StorableConstructorFlag _) : base(_) {
    }

    public override IDeepCloneable Clone(Cloner cloner) {
      return new SelfConfiguration(this, cloner);
    }
    #endregion
    #region dialog functions
    public void Initialization(int size) {
      for (int t = 0; t < size; t++) {
        Probabilities.Add(1.0 / ((double)(size))); // uniform distribution as start point
      }
      SocialKarte = 1.0 / (size * 20.0); // parameters of method
      StepValue = SocialKarte / 5.0;
    }
    public void ReadFromFile(int size, string fileName) {
      SocialKarte = 1.0 / (size * 20.0); // parameters of method
      StepValue = SocialKarte / 5.0;
      var temp = FileComuncations.DoubleMatrixFromFileRead(fileName);
      Probabilities = temp[0];
    }
    public int Aplay(IRandom random) {
      return HelpFunctions.OneElementFromListProportionalSelection(random, Probabilities);
    }
    public int Aplay(int individNumber) {
      return ChoiseForPopulation[individNumber];
    }
    public void DecisionForPopulation(int size, IRandom random) {
      if (ChoiseForPopulation != null) {
        ChoiseForPopulation.Clear();
      }
      for (int i = 0; i < size; i++) {
        ChoiseForPopulation.Add(HelpFunctions.OneElementFromListProportionalSelection(random, Probabilities));
      }
    }
    #region probabilities updating
    public void UpDate(Dictionary<ISymbolicExpressionTree, double> population) {
      SucsessStatisticCollection(population);
      ProbabilitiesUpDate();
    }
    public void UpDate(double[] population) {
      SucsessStatisticCollection(population);
      ProbabilitiesUpDate();
    }
    private void SucsessStatisticCollection(double[] qualities) {
      if (SucsessStatistics.Count != 0)
        SucsessStatistics.Clear();
      for (int t = 0; t < Probabilities.Count; t++) {
        SucsessStatistics.Add(new List<double>());
        SucsessStatistics[t].Add(t);
        SucsessStatistics[t].Add(0);
      }
      for (int i = 0; i < ChoiseForPopulation.Count; i++) {
        SucsessStatistics[ChoiseForPopulation[i]][1] = qualities[i];
      }
    }
    private void SucsessStatisticCollection(Dictionary<ISymbolicExpressionTree, double> population) {
      if (SucsessStatistics.Count != 0)
        SucsessStatistics.Clear();
      for (int t = 0; t < Probabilities.Count; t++) {
        SucsessStatistics.Add(new List<double>());
        SucsessStatistics[t].Add(0);
        SucsessStatistics[t].Add(0);
      }
      foreach (var solution in population) {
        TreeCheck(solution.Key, solution.Value);
      }
    }
    private void TreeCheck(ISymbolicExpressionTree tree, double treeQuality) {
      foreach (var treeNode in tree.IterateNodesPrefix().OfType<TreeModelTreeNode>()) {
        SucsessStatistics[treeNode.TreeNumber][0] += 1;
        SucsessStatistics[treeNode.TreeNumber][1] += treeQuality;
      }
    }
    private void ProbabilitiesUpDate() {
      var averageQuality = new List<double>();
      foreach (var variant in SucsessStatistics) {
        if (variant[0] > 0.005) {
          averageQuality.Add(variant[1] / variant[0]);
        } else { averageQuality.Add(0); }
      }
      int bestModelNumber = HelpFunctions.ChooseMaxElementIndex(averageQuality);
      double totalChangeValue = 0, changeValue = 0;
      for (int i = 0; i < Probabilities.Count; i++) {
        changeValue = CheckSocialKatre(Probabilities[i]);
        totalChangeValue += changeValue;
        Probabilities[i] -= changeValue;
      }
      Probabilities[bestModelNumber] += totalChangeValue;
    }
    private double CheckSocialKatre(double value) {
      if (value > (SocialKarte + StepValue))
        return StepValue;
      else if (value > SocialKarte)
        return (value - SocialKarte);
      else return 0;
    }
    #endregion
    #endregion
  }
}