source: branches/LearningClassifierSystems/HeuristicLab.Algorithms.LearningClassifierSystems/3.3/LearningClassifierSystem.cs @ 9154

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2012 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;
using System.Linq;
using HeuristicLab.Analysis;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.ConditionActionEncoding;
using HeuristicLab.Operators;
using HeuristicLab.Optimization;
using HeuristicLab.Optimization.Operators;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
using HeuristicLab.Random;

namespace HeuristicLab.Algorithms.LearningClassifierSystems {
  /// <summary>
  /// A learning classifier system.
  /// </summary>
  [Item("Learning Classifier System", "A genetic algorithm.")]
  [Creatable("Algorithms")]
  [StorableClass]
  public sealed class LearningClassifierSystem : HeuristicOptimizationEngineAlgorithm, IStorableContent {
    public string Filename { get; set; }

    #region Problem Properties
    public override Type ProblemType {
      get { return typeof(IConditionActionProblem); }
    }
    public new IConditionActionProblem Problem {
      get { return (IConditionActionProblem)base.Problem; }
      set { base.Problem = value; }
    }
    #endregion

    #region Parameter Properties
    private ValueParameter<IntValue> SeedParameter {
      get { return (ValueParameter<IntValue>)Parameters["Seed"]; }
    }
    private ValueParameter<BoolValue> SetSeedRandomlyParameter {
      get { return (ValueParameter<BoolValue>)Parameters["SetSeedRandomly"]; }
    }
    private ValueParameter<BoolValue> CreateInitialPopulationParameter {
      get { return (ValueParameter<BoolValue>)Parameters["CreateInitialPopulation"]; }
    }
    private ValueParameter<IntValue> PopulationSizeParameter {
      get { return (ValueParameter<IntValue>)Parameters["N"]; }
    }
    private ValueParameter<PercentValue> BetaParameter {
      get { return (ValueParameter<PercentValue>)Parameters["Beta"]; }
    }
    private ValueParameter<PercentValue> AlphaParameter {
      get { return (ValueParameter<PercentValue>)Parameters["Alpha"]; }
    }
    private ValueParameter<DoubleValue> ErrorZeroParameter {
      get { return (ValueParameter<DoubleValue>)Parameters["ErrorZero"]; }
    }
    private ValueParameter<DoubleValue> PowerParameter {
      get { return (ValueParameter<DoubleValue>)Parameters["v"]; }
    }
    private ValueParameter<PercentValue> GammaParameter {
      get { return (ValueParameter<PercentValue>)Parameters["Gamma"]; }
    }
    private ValueParameter<PercentValue> CrossoverProbabilityParameter {
      get { return (ValueParameter<PercentValue>)Parameters["CrossoverProbability"]; }
    }
    private ValueParameter<PercentValue> MutationProbabilityParameter {
      get { return (ValueParameter<PercentValue>)Parameters["MutationProbability"]; }
    }
    private ValueParameter<IntValue> ThetaGAParameter {
      get { return (ValueParameter<IntValue>)Parameters["ThetaGA"]; }
    }
    private ValueParameter<IntValue> ThetaDeletionParameter {
      get { return (ValueParameter<IntValue>)Parameters["ThetaDeletion"]; }
    }
    private ValueParameter<IntValue> ThetaSubsumptionParameter {
      get { return (ValueParameter<IntValue>)Parameters["ThetaSubsumption"]; }
    }
    private ValueParameter<PercentValue> DeltaParameter {
      get { return (ValueParameter<PercentValue>)Parameters["Delta"]; }
    }
    private ValueParameter<PercentValue> ExplorationProbabilityParameter {
      get { return (ValueParameter<PercentValue>)Parameters["ExplorationProbability"]; }
    }
    private ValueParameter<BoolValue> DoGASubsumptionParameter {
      get { return (ValueParameter<BoolValue>)Parameters["DoGASubsumption"]; }
    }
    private ValueParameter<BoolValue> DoActionSetSubsumptionParameter {
      get { return (ValueParameter<BoolValue>)Parameters["DoActionSetSubsumption"]; }
    }
    private ValueParameter<MultiAnalyzer> AnalyzerParameter {
      get { return (ValueParameter<MultiAnalyzer>)Parameters["Analyzer"]; }
    }
    private ValueParameter<IntValue> MaxIterationsParameter {
      get { return (ValueParameter<IntValue>)Parameters["MaxIterations"]; }
    }
    #endregion

    #region Properties
    public IntValue Seed {
      get { return SeedParameter.Value; }
      set { SeedParameter.Value = value; }
    }
    public BoolValue SetSeedRandomly {
      get { return SetSeedRandomlyParameter.Value; }
      set { SetSeedRandomlyParameter.Value = value; }
    }
    public BoolValue CreateInitialPopulation {
      get { return CreateInitialPopulationParameter.Value; }
      set { CreateInitialPopulationParameter.Value = value; }
    }
    public IntValue PopulationSize {
      get { return PopulationSizeParameter.Value; }
      set { PopulationSizeParameter.Value = value; }
    }
    public PercentValue Beta {
      get { return BetaParameter.Value; }
      set { BetaParameter.Value = value; }
    }
    public PercentValue Alpha {
      get { return AlphaParameter.Value; }
      set { AlphaParameter.Value = value; }
    }
    public DoubleValue ErrorZero {
      get { return ErrorZeroParameter.Value; }
      set { ErrorZeroParameter.Value = value; }
    }
    public DoubleValue Power {
      get { return PowerParameter.Value; }
      set { PowerParameter.Value = value; }
    }
    public PercentValue Gamma {
      get { return GammaParameter.Value; }
      set { GammaParameter.Value = value; }
    }
    public PercentValue CrossoverProbability {
      get { return CrossoverProbabilityParameter.Value; }
      set { CrossoverProbabilityParameter.Value = value; }
    }
    public PercentValue MutationProbability {
      get { return MutationProbabilityParameter.Value; }
      set { MutationProbabilityParameter.Value = value; }
    }
    public IntValue ThetaGA {
      get { return ThetaGAParameter.Value; }
      set { ThetaGAParameter.Value = value; }
    }
    public IntValue ThetaDeletion {
      get { return ThetaDeletionParameter.Value; }
      set { ThetaDeletionParameter.Value = value; }
    }
    public IntValue ThetaSubsumption {
      get { return ThetaSubsumptionParameter.Value; }
      set { ThetaSubsumptionParameter.Value = value; }
    }
    public PercentValue Delta {
      get { return DeltaParameter.Value; }
      set { DeltaParameter.Value = value; }
    }
    public PercentValue ExplorationProbability {
      get { return ExplorationProbabilityParameter.Value; }
      set { ExplorationProbabilityParameter.Value = value; }
    }
    public BoolValue DoGASubsumption {
      get { return DoGASubsumptionParameter.Value; }
      set { DoGASubsumptionParameter.Value = value; }
    }
    public BoolValue DoActionSetSubsumption {
      get { return DoActionSetSubsumptionParameter.Value; }
      set { DoActionSetSubsumptionParameter.Value = value; }
    }
    public IntValue MaxIterations {
      get { return MaxIterationsParameter.Value; }
      set { MaxIterationsParameter.Value = value; }
    }
    public MultiAnalyzer Analyzer {
      get { return AnalyzerParameter.Value; }
      set { AnalyzerParameter.Value = value; }
    }
    private RandomCreator RandomCreator {
      get { return (RandomCreator)OperatorGraph.InitialOperator; }
    }
    public LearningClassifierSystemMainLoop MainLoop {
      get { return FindMainLoop(RandomCreator.Successor); }
    }
    #endregion

    public LearningClassifierSystem()
      : base() {
      #region Create parameters
      Parameters.Add(new ValueParameter<IntValue>("Seed", "The random seed used to initialize the new pseudo random number generator.", new IntValue(0)));
      Parameters.Add(new ValueParameter<BoolValue>("SetSeedRandomly", "True if the random seed should be set to a random value, otherwise false.", new BoolValue(true)));
      Parameters.Add(new ValueParameter<BoolValue>("CreateInitialPopulation", "Specifies if a population should be created at the beginning of the algorithm.", new BoolValue(false)));
      Parameters.Add(new ValueParameter<IntValue>("N", "Max size of the population of solutions.", new IntValue(100)));
      Parameters.Add(new ValueParameter<PercentValue>("Beta", "Learning rate", new PercentValue(0.1)));
      Parameters.Add(new ValueParameter<PercentValue>("Alpha", "", new PercentValue(0.1)));
      Parameters.Add(new ValueParameter<DoubleValue>("ErrorZero", "The error below which classifiers are considered to have equal accuracy", new DoubleValue(10)));
      Parameters.Add(new ValueParameter<DoubleValue>("v", "Power parameter", new DoubleValue(5)));
      Parameters.Add(new ValueParameter<PercentValue>("Gamma", "Discount factor", new PercentValue(0.71)));
      Parameters.Add(new ValueParameter<PercentValue>("CrossoverProbability", "Probability of crossover", new PercentValue(0.9)));
      Parameters.Add(new ValueParameter<PercentValue>("MutationProbability", "Probability of mutation", new PercentValue(0.05)));
      Parameters.Add(new ValueParameter<IntValue>("ThetaGA", "GA threshold. GA is applied in a set when the average time since the last GA is greater than ThetaGA.", new IntValue(25)));
      Parameters.Add(new ValueParameter<IntValue>("ThetaDeletion", "Deletion threshold. If the experience of a classifier is greater than ThetaDeletion, its fitness may be considered in its probability of deletion.", new IntValue(20)));
      Parameters.Add(new ValueParameter<IntValue>("ThetaSubsumption", "Subsumption threshold. The experience of a classifier must be greater than TheatSubsumption to be able to subsume another classifier.", new IntValue(20)));
      Parameters.Add(new ValueParameter<PercentValue>("Delta", "Delta specifies the fraction of mean fitness in [P] below which the fitness of a classifier may be considered in its probability of deletion", new PercentValue(0.1)));
      Parameters.Add(new ValueParameter<PercentValue>("ExplorationProbability", "Probability of selecting the action uniform randomly", new PercentValue(0.5)));
      Parameters.Add(new ValueParameter<BoolValue>("DoGASubsumption", "Specifies if offsprings are tested for possible logical subsumption by parents.", new BoolValue(true)));
      Parameters.Add(new ValueParameter<BoolValue>("DoActionSetSubsumption", "Specifies if action set is tested for subsuming classifiers.", new BoolValue(true)));
      Parameters.Add(new ValueParameter<MultiAnalyzer>("Analyzer", "The operator used to analyze each generation.", new MultiAnalyzer()));
      Parameters.Add(new ValueParameter<IntValue>("MaxIterations", "The maximum number of iterations.", new IntValue(1000)));
      #endregion

      #region Create operators
      RandomCreator randomCreator = new RandomCreator();

      ResultsCollector resultsCollector = new ResultsCollector();
      LearningClassifierSystemMainLoop mainLoop = new LearningClassifierSystemMainLoop();

      randomCreator.RandomParameter.ActualName = "Random";
      randomCreator.SeedParameter.ActualName = SeedParameter.Name;
      randomCreator.SeedParameter.Value = null;
      randomCreator.SetSeedRandomlyParameter.ActualName = SetSeedRandomlyParameter.Name;
      randomCreator.SetSeedRandomlyParameter.Value = null;

      resultsCollector.ResultsParameter.ActualName = "Results";

      mainLoop.AnalyzerParameter.ActualName = AnalyzerParameter.Name;
      mainLoop.MaxIterationsParameter.ActualName = MaxIterationsParameter.Name;
      #endregion

      #region Create operator graph
      OperatorGraph.InitialOperator = randomCreator;
      randomCreator.Successor = resultsCollector;
      resultsCollector.Successor = mainLoop;
      #endregion

      UpdateAnalyzers();
    }
    private LearningClassifierSystem(LearningClassifierSystem original, Cloner cloner)
      : base(original, cloner) {
    }
    public override IDeepCloneable Clone(Cloner cloner) {
      return new LearningClassifierSystem(this, cloner);
    }
    [StorableConstructor]
    private LearningClassifierSystem(bool deserializing) : base(deserializing) { }

    protected override void OnProblemChanged() {
      if (Problem != null) {
        ParameterizeEvaluator(Problem.Evaluator);
        MainLoop.SetCurrentProblem(Problem);
        UpdateAnalyzers();
      }
      base.OnProblemChanged();
    }
    protected override void Problem_EvaluatorChanged(object sender, EventArgs e) {
      ParameterizeEvaluator(Problem.Evaluator);
      MainLoop.SetCurrentProblem(Problem);
      base.Problem_EvaluatorChanged(sender, e);
    }
    protected override void Problem_SolutionCreatorChanged(object sender, EventArgs e) {
      MainLoop.SetCurrentProblem(Problem);
      base.Problem_SolutionCreatorChanged(sender, e);
    }
    protected override void Problem_OperatorsChanged(object sender, EventArgs e) {
      UpdateAnalyzers();
      base.Problem_OperatorsChanged(sender, e);
    }

    private void ParameterizeEvaluator(IXCSEvaluator evaluator) {
      evaluator.ActualTimeParameter.ActualName = "Iteration";
      evaluator.BetaParameter.ActualName = BetaParameter.Name;
      evaluator.AlphaParameter.ActualName = AlphaParameter.Name;
      evaluator.PowerParameter.ActualName = PowerParameter.Name;
      evaluator.ErrorZeroParameter.ActualName = ErrorZeroParameter.Name;
    }

    private void UpdateAnalyzers() {
      Analyzer.Operators.Clear();
      if (Problem != null) {
        foreach (IAnalyzer analyzer in Problem.Operators.OfType<IAnalyzer>()) {
          Analyzer.Operators.Add(analyzer, analyzer.EnabledByDefault);
        }
      }
    }

    private LearningClassifierSystemMainLoop FindMainLoop(IOperator start) {
      IOperator mainLoop = start;
      while (mainLoop != null && !(mainLoop is LearningClassifierSystemMainLoop))
        mainLoop = ((SingleSuccessorOperator)mainLoop).Successor;
      if (mainLoop == null) return null;
      else return (LearningClassifierSystemMainLoop)mainLoop;
    }
  }
}