1  #region License Information


2  /* HeuristicLab


3  * Copyright (C) 20022018 Heuristic and Evolutionary Algorithms Laboratory (HEAL)


4  *


5  * This file is part of HeuristicLab.


6  *


7  * HeuristicLab is free software: you can redistribute it and/or modify


8  * it under the terms of the GNU General Public License as published by


9  * the Free Software Foundation, either version 3 of the License, or


10  * (at your option) any later version.


11  *


12  * HeuristicLab is distributed in the hope that it will be useful,


13  * but WITHOUT ANY WARRANTY; without even the implied warranty of


14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the


15  * GNU General Public License for more details.


16  *


17  * You should have received a copy of the GNU General Public License


18  * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.


19  */


20  #endregion


21 


22  using System;


23  using System.Collections.Generic;


24  using System.Linq;


25  using HeuristicLab.Common;


26  using HeuristicLab.Core;


27  using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;


28  using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;


29  using HeuristicLab.Random;


30 


31  namespace HeuristicLab.Problems.DataAnalysis.Symbolic {


32  [Item("ProbabilisticFunctionalCrossover", "An operator which performs subtree swapping based on the behavioral similarity between subtrees:\n" +


33  " Take two parent individuals P0 and P1\n" +


34  " Randomly choose a node N from P0\n" +


35  " For each matching node M from P1, calculate the behavioral distance:\n" +


36  "\t\tD(N,M) = 0.5 * ( abs(max(N)  max(M)) + abs(min(N)  min(M)) )\n" +


37  " Make a probabilistic weighted choice of node M from P1, based on the inversed and normalized behavioral distance")]


38  [StorableClass]


39  public sealed class SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover<T> : SymbolicDataAnalysisExpressionCrossover<T> where T : class, IDataAnalysisProblemData {


40  [StorableConstructor]


41  private SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover(bool deserializing) : base(deserializing) { }


42  private SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover(SymbolicDataAnalysisExpressionCrossover<T> original, Cloner cloner)


43  : base(original, cloner) { }


44  public SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover()


45  : base() {


46  name = "ProbabilisticFunctionalCrossover";


47  }


48  public override IDeepCloneable Clone(Cloner cloner) { return new SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover<T>(this, cloner); }


49 


50  public override ISymbolicExpressionTree Crossover(IRandom random, ISymbolicExpressionTree parent0, ISymbolicExpressionTree parent1) {


51  ISymbolicDataAnalysisExpressionTreeInterpreter interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;


52  List<int> rows = GenerateRowsToEvaluate().ToList();


53  T problemData = ProblemDataParameter.ActualValue;


54  return Cross(random, parent0, parent1, interpreter, problemData,


55  rows, MaximumSymbolicExpressionTreeDepth.Value, MaximumSymbolicExpressionTreeLength.Value);


56  }


57 


58  /// <summary>


59  /// Takes two parent individuals P0 and P1.


60  /// Randomly choose a node i from the first parent, then for each matching node j from the second parent, calculate the behavioral distance based on the range:


61  /// d(i,j) = 0.5 * ( abs(max(i)  max(j)) + abs(min(i)  min(j)) ).


62  /// Next, assign probabilities for the selection of a node j based on the inversed and normalized behavioral distance, then make a random weighted choice.


63  /// </summary>


64  public static ISymbolicExpressionTree Cross(IRandom random, ISymbolicExpressionTree parent0, ISymbolicExpressionTree parent1,


65  ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, T problemData, IList<int> rows, int maxDepth, int maxLength) {


66  var crossoverPoints0 = new List<CutPoint>();


67  parent0.Root.ForEachNodePostfix((n) => {


68  // the if clauses prevent the root or the startnode from being selected, although the startnode can be the parent of the node being swapped.


69  if (n.Parent != null && n.Parent != parent0.Root) {


70  crossoverPoints0.Add(new CutPoint(n.Parent, n));


71  }


72  });


73 


74  var crossoverPoint0 = crossoverPoints0.SampleRandom(random);


75  int level = parent0.Root.GetBranchLevel(crossoverPoint0.Child);


76  int length = parent0.Root.GetLength()  crossoverPoint0.Child.GetLength();


77 


78  var allowedBranches = new List<ISymbolicExpressionTreeNode>();


79  parent1.Root.ForEachNodePostfix((n) => {


80  if (n.Parent != null && n.Parent != parent1.Root) {


81  if (n.GetDepth() + level <= maxDepth && n.GetLength() + length <= maxLength && crossoverPoint0.IsMatchingPointType(n))


82  allowedBranches.Add(n);


83  }


84  });


85 


86  if (allowedBranches.Count == 0)


87  return parent0;


88 


89  var dataset = problemData.Dataset;


90 


91  // create symbols in order to improvize an adhoc tree so that the child can be evaluated


92  var rootSymbol = new ProgramRootSymbol();


93  var startSymbol = new StartSymbol();


94  var tree0 = CreateTreeFromNode(random, crossoverPoint0.Child, rootSymbol, startSymbol); // this will change crossoverPoint0.Child.Parent


95  double min0 = 0.0, max0 = 0.0;


96  foreach (double v in interpreter.GetSymbolicExpressionTreeValues(tree0, dataset, rows)) {


97  if (min0 > v) min0 = v;


98  if (max0 < v) max0 = v;


99  }


100  crossoverPoint0.Child.Parent = crossoverPoint0.Parent; // restore correct parent


101 


102  var weights = new List<double>();


103  foreach (var node in allowedBranches) {


104  var parent = node.Parent;


105  var tree1 = CreateTreeFromNode(random, node, rootSymbol, startSymbol);


106  double min1 = 0.0, max1 = 0.0;


107  foreach (double v in interpreter.GetSymbolicExpressionTreeValues(tree1, dataset, rows)) {


108  if (min1 > v) min1 = v;


109  if (max1 < v) max1 = v;


110  }


111  double behavioralDistance = (Math.Abs(min0  min1) + Math.Abs(max0  max1)) / 2; // this can be NaN of Infinity because some trees are crazy like exp(exp(exp(...))), we correct that below


112  weights.Add(behavioralDistance);


113  node.Parent = parent; // restore correct node parent


114  }


115 


116  // remove branches with an infinite or NaN behavioral distance


117  for (int i = weights.Count  1; i >= 0; i) {


118  if (Double.IsNaN(weights[i])  Double.IsInfinity(weights[i])) {


119  weights.RemoveAt(i);


120  allowedBranches.RemoveAt(i);


121  }


122  }


123  // check if there are any allowed branches left


124  if (allowedBranches.Count == 0)


125  return parent0;


126 


127  ISymbolicExpressionTreeNode selectedBranch;


128  double sum = weights.Sum();


129 


130  if (sum.IsAlmost(0.0)  weights.Count == 1) // if there is only one allowed branch, or if all weights are zero


131  selectedBranch = allowedBranches[0];


132  else {


133  for (int i = 0; i != weights.Count; ++i) // normalize and invert values


134  weights[i] = 1  weights[i] / sum;


135 


136  sum = weights.Sum(); // take new sum


137 


138  // compute the probabilities (selection weights)


139  for (int i = 0; i != weights.Count; ++i)


140  weights[i] /= sum;


141 


142  #pragma warning disable 612, 618


143  selectedBranch = allowedBranches.SelectRandom(weights, random);


144  #pragma warning restore 612, 618


145  }


146  Swap(crossoverPoint0, selectedBranch);


147  return parent0;


148  }


149  }


150  }

