source: branches/DataPreprocessing/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Analyzers/SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer.cs @ 11009

Last change on this file since 11009 was 11009, checked in by pfleck, 5 years ago
  • Merged trunk into preprocessing branch.
File size: 10.3 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.Data;
28using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
29using HeuristicLab.Optimization;
30using HeuristicLab.Parameters;
31using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
32
33namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
34  /// <summary>
35  /// An operator that analyzes the validation best symbolic data analysis solution for single objective symbolic data analysis problems.
36  /// </summary>
37  [Item("SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer", "An operator that analyzes the validation best symbolic data analysis solution for single objective symbolic data analysis problems.")]
38  [StorableClass]
39  public abstract class SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer<S, T, U> : SymbolicDataAnalysisSingleObjectiveValidationAnalyzer<T, U>, IIterationBasedOperator
40    where S : class, ISymbolicDataAnalysisSolution
41    where T : class, ISymbolicDataAnalysisSingleObjectiveEvaluator<U>
42    where U : class, IDataAnalysisProblemData {
43    private const string ValidationBestSolutionParameterName = "Best validation solution";
44    private const string ValidationBestSolutionQualityParameterName = "Best validation solution quality";
45    private const string ValidationBestSolutionGenerationParameterName = "Best validation solution generation";
46    private const string UpdateAlwaysParameterName = "Always update best solution";
47    private const string IterationsParameterName = "Iterations";
48    private const string MaximumIterationsParameterName = "Maximum Iterations";
49
50    #region parameter properties
51    public ILookupParameter<S> ValidationBestSolutionParameter {
52      get { return (ILookupParameter<S>)Parameters[ValidationBestSolutionParameterName]; }
53    }
54    public ILookupParameter<DoubleValue> ValidationBestSolutionQualityParameter {
55      get { return (ILookupParameter<DoubleValue>)Parameters[ValidationBestSolutionQualityParameterName]; }
56    }
57    public ILookupParameter<IntValue> ValidationBestSolutionGenerationParameter {
58      get { return (ILookupParameter<IntValue>)Parameters[ValidationBestSolutionGenerationParameterName]; }
59    }
60    public IFixedValueParameter<BoolValue> UpdateAlwaysParameter {
61      get { return (IFixedValueParameter<BoolValue>)Parameters[UpdateAlwaysParameterName]; }
62    }
63    public ILookupParameter<IntValue> IterationsParameter {
64      get { return (ILookupParameter<IntValue>)Parameters[IterationsParameterName]; }
65    }
66    public IValueLookupParameter<IntValue> MaximumIterationsParameter {
67      get { return (IValueLookupParameter<IntValue>)Parameters[MaximumIterationsParameterName]; }
68    }
69    #endregion
70    #region properties
71    public S ValidationBestSolution {
72      get { return ValidationBestSolutionParameter.ActualValue; }
73      set { ValidationBestSolutionParameter.ActualValue = value; }
74    }
75    public DoubleValue ValidationBestSolutionQuality {
76      get { return ValidationBestSolutionQualityParameter.ActualValue; }
77      set { ValidationBestSolutionQualityParameter.ActualValue = value; }
78    }
79    public BoolValue UpdateAlways {
80      get { return UpdateAlwaysParameter.Value; }
81    }
82    #endregion
83
84    [StorableConstructor]
85    protected SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer(bool deserializing) : base(deserializing) { }
86    protected SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer(SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer<S, T, U> original, Cloner cloner) : base(original, cloner) { }
87    public SymbolicDataAnalysisSingleObjectiveValidationBestSolutionAnalyzer()
88      : base() {
89      Parameters.Add(new LookupParameter<S>(ValidationBestSolutionParameterName, "The validation best symbolic data analyis solution."));
90      Parameters.Add(new LookupParameter<DoubleValue>(ValidationBestSolutionQualityParameterName, "The quality of the validation best symbolic data analysis solution."));
91      Parameters.Add(new LookupParameter<IntValue>(ValidationBestSolutionGenerationParameterName, "The generation in which the best validation solution was found."));
92      Parameters.Add(new FixedValueParameter<BoolValue>(UpdateAlwaysParameterName, "Determines if the best validation solution should always be updated regardless of its quality.", new BoolValue(false)));
93      Parameters.Add(new LookupParameter<IntValue>(IterationsParameterName, "The number of performed iterations."));
94      Parameters.Add(new ValueLookupParameter<IntValue>(MaximumIterationsParameterName, "The maximum number of performed iterations.") { Hidden = true });
95      UpdateAlwaysParameter.Hidden = true;
96    }
97
98    [StorableHook(HookType.AfterDeserialization)]
99    private void AfterDeserialization() {
100      if (!Parameters.ContainsKey(UpdateAlwaysParameterName)) {
101        Parameters.Add(new FixedValueParameter<BoolValue>(UpdateAlwaysParameterName, "Determines if the best validation solution should always be updated regardless of its quality.", new BoolValue(false)));
102        UpdateAlwaysParameter.Hidden = true;
103      }
104      if (!Parameters.ContainsKey(ValidationBestSolutionGenerationParameterName))
105        Parameters.Add(new LookupParameter<IntValue>(ValidationBestSolutionGenerationParameterName, "The generation in which the best validation solution was found."));
106      if (!Parameters.ContainsKey(IterationsParameterName))
107        Parameters.Add(new LookupParameter<IntValue>(IterationsParameterName, "The number of performed iterations."));
108      if (!Parameters.ContainsKey(MaximumIterationsParameterName))
109        Parameters.Add(new ValueLookupParameter<IntValue>(MaximumIterationsParameterName, "The maximum number of performed iterations.") { Hidden = true });
110    }
111
112    public override IOperation Apply() {
113      IEnumerable<int> rows = GenerateRowsToEvaluate();
114      if (!rows.Any()) return base.Apply();
115
116      #region find best tree
117      var evaluator = EvaluatorParameter.ActualValue;
118      var problemData = ProblemDataParameter.ActualValue;
119      double bestValidationQuality = Maximization.Value ? double.NegativeInfinity : double.PositiveInfinity;
120      ISymbolicExpressionTree bestTree = null;
121      ISymbolicExpressionTree[] tree = SymbolicExpressionTree.ToArray();
122
123      // sort is ascending and we take the first n% => order so that best solutions are smallest
124      // sort order is determined by maximization parameter
125      double[] trainingQuality;
126      if (Maximization.Value) {
127        // largest values must be sorted first
128        trainingQuality = Quality.Select(x => -x.Value).ToArray();
129      } else {
130        // smallest values must be sorted first
131        trainingQuality = Quality.Select(x => x.Value).ToArray();
132      }
133
134      // sort trees by training qualities
135      Array.Sort(trainingQuality, tree);
136
137      // number of best training solutions to validate (at least 1)
138      int topN = (int)Math.Max(tree.Length * PercentageOfBestSolutionsParameter.ActualValue.Value, 1);
139
140      IExecutionContext childContext = (IExecutionContext)ExecutionContext.CreateChildOperation(evaluator);
141      // evaluate best n training trees on validiation set
142      var quality = tree
143        .Take(topN)
144        .Select(t => evaluator.Evaluate(childContext, t, problemData, rows))
145        .ToArray();
146
147      for (int i = 0; i < quality.Length; i++) {
148        if (IsBetter(quality[i], bestValidationQuality, Maximization.Value)) {
149          bestValidationQuality = quality[i];
150          bestTree = tree[i];
151        }
152      }
153      #endregion
154
155      var results = ResultCollection;
156      if (UpdateAlways.Value || ValidationBestSolutionQuality == null ||
157        IsBetter(bestValidationQuality, ValidationBestSolutionQuality.Value, Maximization.Value)) {
158        ValidationBestSolution = CreateSolution(bestTree, bestValidationQuality);
159        ValidationBestSolutionQuality = new DoubleValue(bestValidationQuality);
160        if (IterationsParameter.ActualValue != null)
161          ValidationBestSolutionGenerationParameter.ActualValue = new IntValue(IterationsParameter.ActualValue.Value);
162
163        if (!results.ContainsKey(ValidationBestSolutionParameter.Name)) {
164          results.Add(new Result(ValidationBestSolutionParameter.Name, ValidationBestSolutionParameter.Description, ValidationBestSolution));
165          results.Add(new Result(ValidationBestSolutionQualityParameter.Name, ValidationBestSolutionQualityParameter.Description, ValidationBestSolutionQuality));
166          if (ValidationBestSolutionGenerationParameter.ActualValue != null)
167            results.Add(new Result(ValidationBestSolutionGenerationParameter.Name, ValidationBestSolutionGenerationParameter.Description, ValidationBestSolutionGenerationParameter.ActualValue));
168        } else {
169          results[ValidationBestSolutionParameter.Name].Value = ValidationBestSolution;
170          results[ValidationBestSolutionQualityParameter.Name].Value = ValidationBestSolutionQuality;
171          if (ValidationBestSolutionGenerationParameter.ActualValue != null)
172            results[ValidationBestSolutionGenerationParameter.Name].Value = ValidationBestSolutionGenerationParameter.ActualValue;
173        }
174      }
175      return base.Apply();
176    }
177
178    protected abstract S CreateSolution(ISymbolicExpressionTree bestTree, double bestQuality);
179
180    private bool IsBetter(double lhs, double rhs, bool maximization) {
181      if (maximization) return lhs > rhs;
182      else return lhs < rhs;
183    }
184  }
185}
Note: See TracBrowser for help on using the repository browser.