Free cookie consent management tool by TermsFeed Policy Generator

source: branches/Sliding Window GP/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Crossovers/SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover.cs @ 9708

Last change on this file since 9708 was 9708, checked in by mkommend, 11 years ago

#1837: Merged trunk changes into sliding window gp branch.

File size: 7.5 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2013 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
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Core;
27using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
28using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
29
30namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
31  [Item("ProbabilisticFunctionalCrossover", "An operator which performs subtree swapping based on the behavioral similarity between subtrees:\n" +
32                                            "- Take two parent individuals P0 and P1\n" +
33                                            "- Randomly choose a node N from P0\n" +
34                                            "- For each matching node M from P1, calculate the behavioral distance:\n" +
35                                            "\t\tD(N,M) = 0.5 * ( abs(max(N) - max(M)) + abs(min(N) - min(M)) )\n" +
36                                            "- Make a probabilistic weighted choice of node M from P1, based on the inversed and normalized behavioral distance")]
37  public sealed class SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover<T> : SymbolicDataAnalysisExpressionCrossover<T> where T : class, IDataAnalysisProblemData {
38    [StorableConstructor]
39    private SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover(bool deserializing) : base(deserializing) { }
40    private SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover(SymbolicDataAnalysisExpressionCrossover<T> original, Cloner cloner)
41      : base(original, cloner) { }
42    public SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover()
43      : base() {
44      name = "ProbabilisticFunctionalCrossover";
45    }
46    public override IDeepCloneable Clone(Cloner cloner) { return new SymbolicDataAnalysisExpressionProbabilisticFunctionalCrossover<T>(this, cloner); }
47
48    public override ISymbolicExpressionTree Crossover(IRandom random, ISymbolicExpressionTree parent0, ISymbolicExpressionTree parent1) {
49      ISymbolicDataAnalysisExpressionTreeInterpreter interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
50      List<int> rows = GenerateRowsToEvaluate().ToList();
51      T problemData = ProblemDataParameter.ActualValue;
52      return Cross(random, parent0, parent1, interpreter, problemData,
53                   rows, MaximumSymbolicExpressionTreeDepth.Value, MaximumSymbolicExpressionTreeLength.Value);
54    }
55
56    /// <summary>
57    /// Takes two parent individuals P0 and P1.
58    /// 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:
59    /// d(i,j) = 0.5 * ( abs(max(i) - max(j)) + abs(min(i) - min(j)) ).
60    /// Next, assign probabilities for the selection of a node j based on the inversed and normalized behavioral distance, then make a random weighted choice.
61    /// </summary>
62    public static ISymbolicExpressionTree Cross(IRandom random, ISymbolicExpressionTree parent0, ISymbolicExpressionTree parent1,
63                                                ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, T problemData, IList<int> rows, int maxDepth, int maxLength) {
64      var crossoverPoints0 = new List<CutPoint>();
65      parent0.Root.ForEachNodePostfix((n) => {
66        // the if clauses prevent the root or the startnode from being selected, although the startnode can be the parent of the node being swapped.
67        if (n.Parent != null && n.Parent != parent0.Root) {
68          crossoverPoints0.Add(new CutPoint(n.Parent, n));
69        }
70      });
71      var crossoverPoint0 = crossoverPoints0.SelectRandom(random);
72      int level = parent0.Root.GetBranchLevel(crossoverPoint0.Child);
73      int length = parent0.Root.GetLength() - crossoverPoint0.Child.GetLength();
74
75      var allowedBranches = new List<ISymbolicExpressionTreeNode>();
76      parent1.Root.ForEachNodePostfix((n) => {
77        if (n.Parent != null && n.Parent != parent1.Root) {
78          if (n.GetDepth() + level <= maxDepth && n.GetLength() + length <= maxLength && crossoverPoint0.IsMatchingPointType(n))
79            allowedBranches.Add(n);
80        }
81      });
82
83      if (allowedBranches.Count == 0)
84        return parent0;
85
86      var dataset = problemData.Dataset;
87
88      // create symbols in order to improvize an ad-hoc tree so that the child can be evaluated
89      var rootSymbol = new ProgramRootSymbol();
90      var startSymbol = new StartSymbol();
91      var tree0 = CreateTreeFromNode(random, crossoverPoint0.Child, rootSymbol, startSymbol); // this will change crossoverPoint0.Child.Parent
92      double min0 = 0.0, max0 = 0.0;
93      foreach (double v in interpreter.GetSymbolicExpressionTreeValues(tree0, dataset, rows)) {
94        if (min0 > v) min0 = v;
95        if (max0 < v) max0 = v;
96      }
97      crossoverPoint0.Child.Parent = crossoverPoint0.Parent; // restore correct parent
98
99      var weights = new List<double>();
100      foreach (var node in allowedBranches) {
101        var parent = node.Parent;
102        var tree1 = CreateTreeFromNode(random, node, rootSymbol, startSymbol);
103        double min1 = 0.0, max1 = 0.0;
104        foreach (double v in interpreter.GetSymbolicExpressionTreeValues(tree1, dataset, rows)) {
105          if (min1 > v) min1 = v;
106          if (max1 < v) max1 = v;
107        }
108        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
109        weights.Add(behavioralDistance);
110        node.Parent = parent; // restore correct node parent
111      }
112
113      // remove branches with an infinite or NaN behavioral distance
114      for (int i = weights.Count - 1; i >= 0; --i) {
115        if (Double.IsNaN(weights[i]) || Double.IsInfinity(weights[i])) {
116          weights.RemoveAt(i);
117          allowedBranches.RemoveAt(i);
118        }
119      }
120      // check if there are any allowed branches left
121      if (allowedBranches.Count == 0)
122        return parent0;
123
124      ISymbolicExpressionTreeNode selectedBranch;
125      double sum = weights.Sum();
126
127      if (sum.IsAlmost(0.0) || weights.Count == 1) // if there is only one allowed branch, or if all weights are zero
128        selectedBranch = allowedBranches[0];
129      else {
130        for (int i = 0; i != weights.Count; ++i) // normalize and invert values
131          weights[i] = 1 - weights[i] / sum;
132
133        sum = weights.Sum(); // take new sum
134
135        // compute the probabilities (selection weights)
136        for (int i = 0; i != weights.Count; ++i)
137          weights[i] /= sum;
138
139        selectedBranch = allowedBranches.SelectRandom(weights, random);
140      }
141      Swap(crossoverPoint0, selectedBranch);
142      return parent0;
143    }
144  }
145}
Note: See TracBrowser for help on using the repository browser.