source: branches/DataAnalysis/HeuristicLab.Problems.DataAnalysis/3.3/Operators/DynOpEqComparator.cs @ 4227

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2010 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 alglib;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Operators;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using System.Collections.Generic;

namespace HeuristicLab.Problems.DataAnalysis.Operators {
  [Item("DynOpEqComparator", "Dynamic Operator Equalization.")]
  [StorableClass]
  public class DynOpEqComparator : SingleSuccessorOperator, ISubScopesQualityComparator {
    public IValueLookupParameter<BoolValue> MaximizationParameter {
      get { return (IValueLookupParameter<BoolValue>)Parameters["Maximization"]; }
    }
    public ILookupParameter<BoolValue> ResultParameter {
      get { return (ILookupParameter<BoolValue>)Parameters["Result"]; }
    }
    public ILookupParameter<DoubleValue> LeftSideParameter {
      get { return (ILookupParameter<DoubleValue>)Parameters["LeftSide"]; }
    }
    public ILookupParameter<ItemArray<DoubleValue>> RightSideParameter {
      get { return (ILookupParameter<ItemArray<DoubleValue>>)Parameters["RightSide"]; }
    }
    public ILookupParameter<SymbolicExpressionTree> SymbolicExpressionTreeParameter {
      get { return (ILookupParameter<SymbolicExpressionTree>)Parameters["SymbolicExpressionTree"]; }
    }
    public ILookupParameter<DoubleValue> BestQualityParameter {
      get { return (ILookupParameter<DoubleValue>)Parameters["BestQuality"]; }
    }
    public IValueLookupParameter<IntValue> BinSizeParameter {
      get { return (IValueLookupParameter<IntValue>)Parameters["BinSize"]; }
    }
    public ILookupParameter<ItemList<IntValue>> BinCapacityParameter {
      get { return (ILookupParameter<ItemList<IntValue>>)Parameters["BinCapacity"]; }
    }
    public ILookupParameter<ItemList<ItemList<DoubleValue>>> AcceptedBinQualitiesParameter {
      get { return (ILookupParameter<ItemList<ItemList<DoubleValue>>>)Parameters["AcceptedBinQualities"]; }
    }
    public ILookupParameter<ItemList<IntValue>> AcceptedCountsParameter {
      get { return (ILookupParameter<ItemList<IntValue>>)Parameters["AcceptedCounts"]; }
    }
    public ILookupParameter<ItemList<IntValue>> TotalCountsParameter {
      get { return (ILookupParameter<ItemList<IntValue>>)Parameters["TotalCounts"]; }
    }
    public IValueLookupParameter<BoolValue> AntiOverfitParameter {
      get { return (IValueLookupParameter<BoolValue>)Parameters["AntiOverfit"]; }
    }
    public ILookupParameter<DoubleValue> CurrentBestValidationQualityParameter {
      get { return (ILookupParameter<DoubleValue>)Parameters["Current best validation quality"]; }
    }
    public ILookupParameter<DoubleValue> BestValidationQualityParameter {
      get { return (ILookupParameter<DoubleValue>)Parameters["Best solution quality (validation)"]; }
    }
    public IValueLookupParameter<DoubleValue> CRaiseParameter {
      get { return (IValueLookupParameter<DoubleValue>)Parameters["CRaise"]; }
    }

    public DynOpEqComparator()
      : base() {
      Parameters.Add(new ValueLookupParameter<BoolValue>("Maximization", "True if the problem is a maximization problem, false otherwise"));
      Parameters.Add(new LookupParameter<BoolValue>("Result", "The result of the comparison: True means Quality is better, False means it is worse than parents."));
      Parameters.Add(new LookupParameter<DoubleValue>("LeftSide", "The quality of the child."));
      Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("RightSide", "The qualities of the parents."));
      Parameters.Add(new LookupParameter<SymbolicExpressionTree>("SymbolicExpressionTree", "The newly created child."));
      Parameters.Add(new LookupParameter<DoubleValue>("BestQuality"));
      Parameters.Add(new ValueLookupParameter<IntValue>("BinSize", new IntValue(5)));
      Parameters.Add(new LookupParameter<ItemList<IntValue>>("BinCapacity"));
      Parameters.Add(new LookupParameter<ItemList<ItemList<DoubleValue>>>("AcceptedBinQualities"));
      Parameters.Add(new LookupParameter<ItemList<IntValue>>("AcceptedCounts"));
      Parameters.Add(new LookupParameter<ItemList<IntValue>>("TotalCounts"));
      Parameters.Add(new ValueLookupParameter<BoolValue>("AntiOverfit", new BoolValue(false)));
      Parameters.Add(new LookupParameter<DoubleValue>("Current best validation quality"));
      Parameters.Add(new LookupParameter<DoubleValue>("Best solution quality (validation)"));
      Parameters.Add(new ValueLookupParameter<DoubleValue>("CRaise", "Necessary quality improvement per bin to allow extensions.", new DoubleValue(0.005)));
    }

    public override IOperation Apply() {
      if (ResultParameter.ActualValue == null || ResultParameter.ActualValue.Value == true) {
        var tree = SymbolicExpressionTreeParameter.ActualValue;
        int size = tree.Size;
        int bin = GetBinIndexForSize(size);
        if (LeftSideParameter.ActualValue == null) {
          // not yet evaluated
          #region debugging
          ItemList<IntValue> totalCounts = TotalCountsParameter.ActualValue;
          while (bin >= totalCounts.Count) totalCounts.Add(new IntValue(0));
          totalCounts[bin].Value = totalCounts[bin].Value + 1;
          #endregion
          if (!Exists(bin)) {
            if (AntiOverfitParameter.ActualValue.Value) {
              // reject more complex solutions if the current validation quality is worse than the best so far
              ResultParameter.ActualValue = new BoolValue(!IsOverfitting());
            } else {
              // new bin -> evaluate and check later
              ResultParameter.ActualValue = new BoolValue(true);
            }
          } else {
            // bin exists: 
            // if bin is full -> reject
            // otherwise -> evaluate and check success criterion
            ResultParameter.ActualValue = new BoolValue(IsNotFull(bin));
          }
        } else {
          double leftQuality = LeftSideParameter.ActualValue.Value;
          ResultParameter.ActualValue = new BoolValue(Accept(size, bin, leftQuality));
        }
      }
      return base.Apply();
    }

    private bool IsOverfitting() {
      bool maximization = MaximizationParameter.ActualValue.Value;
      if (CurrentBestValidationQualityParameter.ActualValue != null && BestValidationQualityParameter.ActualValue != null) {
        double currentValidationQuality = CurrentBestValidationQualityParameter.ActualValue.Value;
        double bestValidationQuality = BestValidationQualityParameter.ActualValue.Value;
        return maximization ? currentValidationQuality < bestValidationQuality : currentValidationQuality > bestValidationQuality;
      } else
        return false;
    }

    private int GetBinIndexForSize(int size) {
      return (int)Math.Floor((size - 3.0) / BinSizeParameter.ActualValue.Value);
    }

    private bool Accept(int size, int binIndex, double solutionQuality) {
      bool accept = false;
      if (Exists(binIndex)) {
        //if (IsNotFull(binIndex) /*||
        //NewBestOfBin(solutionQuality, binIndex)*/) {        
        AddToBin(solutionQuality, binIndex);
        accept = true;
        UpdateBestQuality(solutionQuality);
        //}
      } else if (NewBestOfRun(solutionQuality) && SignificantImprovement(solutionQuality, binIndex)) {
        CreateNewBin(binIndex);
        AddToBin(solutionQuality, binIndex);
        accept = true;
        UpdateBestQuality(solutionQuality);
      }
      return accept;
    }

    private void UpdateBestQuality(double solutionQuality) {
      bool maximization = MaximizationParameter.ActualValue.Value;
      double bestQuality = BestQualityParameter.ActualValue.Value;
      if ((maximization && bestQuality < solutionQuality) ||
         (!maximization && solutionQuality < bestQuality))
        BestQualityParameter.ActualValue.Value = solutionQuality;
    }

    private bool SignificantImprovement(double solutionQuality, int newIndex) {
      var binCapacities = BinCapacityParameter.ActualValue;
      int binDiff = newIndex - binCapacities.Count + 1;
      double bestQuality = BestQualityParameter.ActualValue.Value;
      bool maximization = MaximizationParameter.ActualValue.Value;

      double relativeQuality = maximization ? solutionQuality / bestQuality - 1 : bestQuality / solutionQuality - 1;
      return relativeQuality >= binDiff * CRaiseParameter.ActualValue.Value;
    }

    private void AddToBin(double solutionQuality, int binIndex) {
      ItemList<DoubleValue> acceptedBinQualities = AcceptedBinQualitiesParameter.ActualValue[binIndex];
      ItemList<IntValue> acceptedCounts = AcceptedCountsParameter.ActualValue;
      acceptedBinQualities.Add(new DoubleValue(solutionQuality));
      acceptedCounts[binIndex].Value = acceptedCounts[binIndex].Value + 1;
    }

    private bool NewBestOfRun(double solutionQuality) {
      bool maximization = MaximizationParameter.ActualValue.Value;
      double bestQuality = BestQualityParameter.ActualValue.Value;
      return maximization ? solutionQuality > bestQuality : solutionQuality < bestQuality;
    }

    private void CreateNewBin(int binIndex) {
      ItemList<IntValue> binCapacities = BinCapacityParameter.ActualValue;
      ItemList<ItemList<DoubleValue>> acceptedQualities = AcceptedBinQualitiesParameter.ActualValue;
      ItemList<IntValue> acceptedCounts = AcceptedCountsParameter.ActualValue;

      // create empty bins of capacity one up to the newly created bin
      for (int i = binCapacities.Count; i <= binIndex; i++) {
        binCapacities.Add(new IntValue(1));
        acceptedQualities.Add(new ItemList<DoubleValue>(10));
        acceptedCounts.Add(new IntValue(0));
      }
    }

    //private bool NewBestOfBin(double solutionQuality, int binIndex) {
    //  ItemList<ItemList<DoubleValue>> acceptedQualities = AcceptedBinQualitiesParameter.ActualValue;
    //  if (acceptedQualities[binIndex].Count == 0) return true;
    //  bool maximization = MaximizationParameter.ActualValue.Value;
    //  IEnumerable<double> binQualities = acceptedQualities[binIndex].Select(x => x.Value);
    //  // binQualities are always sorted so that the best is in bin 0
    //  return maximization ? solutionQuality > binQualities.First() :
    //    solutionQuality < binQualities.First();
    //}

    private bool IsNotFull(int binIndex) {
      ItemList<IntValue> binCapacities = BinCapacityParameter.ActualValue;
      ItemList<ItemList<DoubleValue>> acceptedQualities = AcceptedBinQualitiesParameter.ActualValue;
      return acceptedQualities[binIndex].Count < binCapacities[binIndex].Value;
    }

    private bool Exists(int binIndex) {
      // if the bin has a capacity set then it exists
      ItemList<IntValue> binCapacities = BinCapacityParameter.ActualValue;
      return binIndex < binCapacities.Count;
    }
  }
}