source: branches/HeuristicLab.EvolutionTracking/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Analyzers/SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer.cs @ 12892

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2015 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
 *
 * This file is part of HeuristicLab.
 *
 * HeuristicLab is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * HeuristicLab is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
 */
#endregion

using System.Linq;
using HeuristicLab.Analysis;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using HeuristicLab.EvolutionTracking;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;

namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Analyzers {
  [StorableClass]
  [Item("SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer", "An analyzer which records the best and average genetic operator improvement")]
  public class SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer : EvolutionTrackingAnalyzer<ISymbolicExpressionTree> {
    public const string QualityParameterName = "Quality";
    public const string PopulationParameterName = "SymbolicExpressionTree";
    public const string CountIntermediateChildrenParameterName = "CountIntermediateChildren";

    public IScopeTreeLookupParameter<DoubleValue> QualityParameter {
      get { return (IScopeTreeLookupParameter<DoubleValue>)Parameters[QualityParameterName]; }
    }

    public IScopeTreeLookupParameter<ISymbolicExpressionTree> PopulationParameter {
      get { return (IScopeTreeLookupParameter<ISymbolicExpressionTree>)Parameters[PopulationParameterName]; }
    }

    public IFixedValueParameter<BoolValue> CountIntermediateChildrenParameter {
      get { return (IFixedValueParameter<BoolValue>)Parameters[CountIntermediateChildrenParameterName]; }
    }

    public bool CountIntermediateChildren {
      get { return CountIntermediateChildrenParameter.Value.Value; }
      set { CountIntermediateChildrenParameter.Value.Value = value; }
    }

    public SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer() {
      Parameters.Add(new ScopeTreeLookupParameter<ISymbolicExpressionTree>(PopulationParameterName, "The population of individuals."));
      Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>(QualityParameterName, "The individual qualities."));
      Parameters.Add(new FixedValueParameter<BoolValue>(CountIntermediateChildrenParameterName, "Specifies whether to consider intermediate children (when crossover was followed by mutation). This should be set to false for offspring selection.", new BoolValue(true)));

      CountIntermediateChildrenParameter.Hidden = true;
    }

    public SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer(
    SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer original, Cloner cloner) : base(original, cloner) {
      CountIntermediateChildren = original.CountIntermediateChildren;
    }

    public override IDeepCloneable Clone(Cloner cloner) {
      return new SymbolicDataAnalysisGeneticOperatorImprovementAnalyzer(this, cloner);
    }

    [StorableHook(HookType.AfterDeserialization)]
    private void AfterDeserialization() {
      if (!Parameters.ContainsKey(CountIntermediateChildrenParameterName))
        Parameters.Add(new FixedValueParameter<BoolValue>(CountIntermediateChildrenParameterName, "Specifies whether to consider intermediate children (when crossover was followed by mutation", new BoolValue(true)));
      CountIntermediateChildrenParameter.Hidden = true;
    }

    public override void ClearState() {
      base.ClearState();
      UpdateCounter.Value = 0;
    }

    public override IOperation Apply() {
      int updateInterval = UpdateIntervalParameter.Value.Value;
      IntValue updateCounter = UpdateCounterParameter.ActualValue;
      if (updateCounter == null) {
        updateCounter = new IntValue(updateInterval);
        UpdateCounterParameter.ActualValue = updateCounter;
      } else updateCounter.Value++;

      if (updateCounter.Value == updateInterval) {
        updateCounter.Value = 0;

        var graph = PopulationGraph;
        if (graph == null || Generation.Value == 0)
          return base.Apply();

        var generation = Generation.Value;
        var averageQuality = QualityParameter.ActualValue.Average(x => x.Value);
        var population = PopulationParameter.ActualValue;
        var populationSize = population.Length;

        var vertices = population.Select(graph.GetByContent).ToList();
        DataTable table;
        double aac = 0; // ratio of above average children produced
        double aacp = 0; // ratio of above average children from above average parents
        #region crossover improvement
        if (!Results.ContainsKey("Crossover improvement")) {
          table = new DataTable("Crossover improvement");
          Results.Add(new Result("Crossover improvement", table));
          table.Rows.AddRange(
            new[]
            {
              new DataRow("Average crossover improvement (root parent)") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Average crossover improvement (non-root parent)") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Average child-parents quality difference") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Best crossover improvement (root parent)") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Best crossover improvement (non-root parent)") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Above average children") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Above average children from above average parents") {VisualProperties = {StartIndexZero = true}},
            });
        } else {
          table = (DataTable)Results["Crossover improvement"].Value;
        }
        var crossoverChildren = vertices.Where(x => x.InDegree == 2).ToList();
        if (CountIntermediateChildren)
          crossoverChildren.AddRange(vertices.Where(x => x.InDegree == 1).Select(v => v.Parents.First()).Where(p => p.Rank.IsAlmost(generation - 0.5))); // add intermediate children

        foreach (var c in crossoverChildren) {
          if (c.Quality > averageQuality) {
            aac++;
            if (c.Parents.All(x => x.Quality > averageQuality))
              aacp++;
          }
        }
        var avgRootParentQualityImprovement = crossoverChildren.Average(x => x.Quality - x.Parents.First().Quality);
        var avgNonRootParentQualityImprovement = crossoverChildren.Average(x => x.Quality - x.Parents.Last().Quality);
        var avgChildParentQuality = crossoverChildren.Average(x => x.Quality - x.Parents.Average(p => p.Quality));
        var bestRootParentQualityImprovement = crossoverChildren.Max(x => x.Quality - x.Parents.First().Quality);
        var bestNonRootParentQualityImprovement = crossoverChildren.Max(x => x.Quality - x.Parents.Last().Quality);
        table.Rows["Average crossover improvement (root parent)"].Values.Add(avgRootParentQualityImprovement);
        table.Rows["Average crossover improvement (non-root parent)"].Values.Add(avgNonRootParentQualityImprovement);
        table.Rows["Best crossover improvement (root parent)"].Values.Add(bestRootParentQualityImprovement);
        table.Rows["Best crossover improvement (non-root parent)"].Values.Add(bestNonRootParentQualityImprovement);
        table.Rows["Average child-parents quality difference"].Values.Add(avgChildParentQuality);
        table.Rows["Above average children"].Values.Add(aac / populationSize);
        table.Rows["Above average children from above average parents"].Values.Add(aacp / populationSize);
        #endregion

        #region mutation improvement
        if (!Results.ContainsKey("Mutation improvement")) {
          table = new DataTable("Mutation improvement");
          Results.Add(new Result("Mutation improvement", table));
          table.Rows.AddRange(
            new[]
            {
              new DataRow("Average mutation improvement") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Best mutation improvement") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Above average children") {VisualProperties = {StartIndexZero = true}},
              new DataRow("Above average children from above average parents") {VisualProperties = {StartIndexZero = true}},
            });
        } else {
          table = (DataTable)Results["Mutation improvement"].Value;
        }

        aac = 0;
        aacp = 0;
        var mutationChildren = vertices.Where(x => x.InDegree == 1).ToList();

        foreach (var c in mutationChildren) {
          if (c.Quality > averageQuality)
            aac++;
          if (c.Parents.All(x => x.Quality > averageQuality))
            aacp++;
        }
        var avgMutationImprovement = mutationChildren.Average(x => x.Quality - x.Parents.First().Quality);
        var bestMutationImprovement = mutationChildren.Max(x => x.Quality - x.Parents.First().Quality);

        table.Rows["Average mutation improvement"].Values.Add(avgMutationImprovement);
        table.Rows["Best mutation improvement"].Values.Add(bestMutationImprovement);
        table.Rows["Above average children"].Values.Add(aac / populationSize);
        table.Rows["Above average children from above average parents"].Values.Add(aacp / populationSize);
        #endregion
      }
      return base.Apply();
    }
  }
}