Free cookie consent management tool by TermsFeed Policy Generator

source: branches/ScatterSearch (trunk integration)/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/SymbolicDataAnalysisProblem.cs @ 7887

Last change on this file since 7887 was 7726, checked in by gkronber, 13 years ago

#1823

  • added analyzer that calculates the complexity of symbolic data analysis trees (weighted symbols)
  • added analyzer that collects the Pareto-optimal solutions regarding complexity and accuracy
File size: 18.4 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2012 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.Drawing;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Common.Resources;
27using HeuristicLab.Core;
28using HeuristicLab.Data;
29using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
30using HeuristicLab.Optimization;
31using HeuristicLab.Parameters;
32using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
33using HeuristicLab.PluginInfrastructure;
34
35namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
36  [StorableClass]
37  public abstract class SymbolicDataAnalysisProblem<T, U, V> : HeuristicOptimizationProblem<U, V>, IDataAnalysisProblem<T>, ISymbolicDataAnalysisProblem, IStorableContent
38    where T : class, IDataAnalysisProblemData
39    where U : class, ISymbolicDataAnalysisEvaluator<T>
40    where V : class, ISymbolicDataAnalysisSolutionCreator {
41
42    #region parameter names & descriptions
43    private const string ProblemDataParameterName = "ProblemData";
44    private const string SymbolicExpressionTreeGrammarParameterName = "SymbolicExpressionTreeGrammar";
45    private const string SymbolicExpressionTreeInterpreterParameterName = "SymbolicExpressionTreeInterpreter";
46    private const string MaximumSymbolicExpressionTreeDepthParameterName = "MaximumSymbolicExpressionTreeDepth";
47    private const string MaximumSymbolicExpressionTreeLengthParameterName = "MaximumSymbolicExpressionTreeLength";
48    private const string MaximumFunctionDefinitionsParameterName = "MaximumFunctionDefinitions";
49    private const string MaximumFunctionArgumentsParameterName = "MaximumFunctionArguments";
50    private const string RelativeNumberOfEvaluatedSamplesParameterName = "RelativeNumberOfEvaluatedSamples";
51    private const string FitnessCalculationPartitionParameterName = "FitnessCalculationPartition";
52    private const string ValidationPartitionParameterName = "ValidationPartition";
53
54    private const string ProblemDataParameterDescription = "";
55    private const string SymbolicExpressionTreeGrammarParameterDescription = "The grammar that should be used for symbolic expression tree.";
56    private const string SymoblicExpressionTreeInterpreterParameterDescription = "The interpreter that should be used to evaluate the symbolic expression tree.";
57    private const string MaximumSymbolicExpressionTreeDepthParameterDescription = "Maximal depth of the symbolic expression. The minimum depth needed for the algorithm is 3 because two levels are reserved for the ProgramRoot and the Start symbol.";
58    private const string MaximumSymbolicExpressionTreeLengthParameterDescription = "Maximal length of the symbolic expression.";
59    private const string MaximumFunctionDefinitionsParameterDescription = "Maximal number of automatically defined functions";
60    private const string MaximumFunctionArgumentsParameterDescription = "Maximal number of arguments of automatically defined functions.";
61    private const string RelativeNumberOfEvaluatedSamplesParameterDescription = "The relative number of samples of the dataset partition, which should be randomly chosen for evaluation.";
62    private const string FitnessCalculationPartitionParameterDescription = "The partition of the problem data training partition, that should be used to calculate the fitness of an individual.";
63    private const string ValidationPartitionParameterDescription = "The partition of the problem data training partition, that should be used to select the best model from (optional).";
64    #endregion
65
66    #region parameter properties
67    IParameter IDataAnalysisProblem.ProblemDataParameter {
68      get { return ProblemDataParameter; }
69    }
70    public IValueParameter<T> ProblemDataParameter {
71      get { return (IValueParameter<T>)Parameters[ProblemDataParameterName]; }
72    }
73    public IValueParameter<ISymbolicDataAnalysisGrammar> SymbolicExpressionTreeGrammarParameter {
74      get { return (IValueParameter<ISymbolicDataAnalysisGrammar>)Parameters[SymbolicExpressionTreeGrammarParameterName]; }
75    }
76    public IValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter> SymbolicExpressionTreeInterpreterParameter {
77      get { return (IValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>)Parameters[SymbolicExpressionTreeInterpreterParameterName]; }
78    }
79    public IFixedValueParameter<IntValue> MaximumSymbolicExpressionTreeDepthParameter {
80      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumSymbolicExpressionTreeDepthParameterName]; }
81    }
82    public IFixedValueParameter<IntValue> MaximumSymbolicExpressionTreeLengthParameter {
83      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumSymbolicExpressionTreeLengthParameterName]; }
84    }
85    public IFixedValueParameter<IntValue> MaximumFunctionDefinitionsParameter {
86      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumFunctionDefinitionsParameterName]; }
87    }
88    public IFixedValueParameter<IntValue> MaximumFunctionArgumentsParameter {
89      get { return (IFixedValueParameter<IntValue>)Parameters[MaximumFunctionArgumentsParameterName]; }
90    }
91    public IFixedValueParameter<PercentValue> RelativeNumberOfEvaluatedSamplesParameter {
92      get { return (IFixedValueParameter<PercentValue>)Parameters[RelativeNumberOfEvaluatedSamplesParameterName]; }
93    }
94    public IFixedValueParameter<IntRange> FitnessCalculationPartitionParameter {
95      get { return (IFixedValueParameter<IntRange>)Parameters[FitnessCalculationPartitionParameterName]; }
96    }
97    public IFixedValueParameter<IntRange> ValidationPartitionParameter {
98      get { return (IFixedValueParameter<IntRange>)Parameters[ValidationPartitionParameterName]; }
99    }
100    #endregion
101
102    #region properties
103    public string Filename { get; set; }
104    public static new Image StaticItemImage { get { return VSImageLibrary.Type; } }
105
106    IDataAnalysisProblemData IDataAnalysisProblem.ProblemData {
107      get { return ProblemData; }
108    }
109    public T ProblemData {
110      get { return ProblemDataParameter.Value; }
111      set { ProblemDataParameter.Value = value; }
112    }
113
114    public ISymbolicDataAnalysisGrammar SymbolicExpressionTreeGrammar {
115      get { return SymbolicExpressionTreeGrammarParameter.Value; }
116      set { SymbolicExpressionTreeGrammarParameter.Value = value; }
117    }
118    public ISymbolicDataAnalysisExpressionTreeInterpreter SymbolicExpressionTreeInterpreter {
119      get { return SymbolicExpressionTreeInterpreterParameter.Value; }
120      set { SymbolicExpressionTreeInterpreterParameter.Value = value; }
121    }
122
123    public IntValue MaximumSymbolicExpressionTreeDepth {
124      get { return MaximumSymbolicExpressionTreeDepthParameter.Value; }
125    }
126    public IntValue MaximumSymbolicExpressionTreeLength {
127      get { return MaximumSymbolicExpressionTreeLengthParameter.Value; }
128    }
129    public IntValue MaximumFunctionDefinitions {
130      get { return MaximumFunctionDefinitionsParameter.Value; }
131    }
132    public IntValue MaximumFunctionArguments {
133      get { return MaximumFunctionArgumentsParameter.Value; }
134    }
135    public PercentValue RelativeNumberOfEvaluatedSamples {
136      get { return RelativeNumberOfEvaluatedSamplesParameter.Value; }
137    }
138
139    public IntRange FitnessCalculationPartition {
140      get { return FitnessCalculationPartitionParameter.Value; }
141    }
142    public IntRange ValidationPartition {
143      get { return ValidationPartitionParameter.Value; }
144    }
145    #endregion
146
147    [StorableConstructor]
148    protected SymbolicDataAnalysisProblem(bool deserializing) : base(deserializing) { }
149    [StorableHook(HookType.AfterDeserialization)]
150    private void AfterDeserialization() {
151      RegisterEventHandlers();
152    }
153    protected SymbolicDataAnalysisProblem(SymbolicDataAnalysisProblem<T, U, V> original, Cloner cloner)
154      : base(original, cloner) {
155      RegisterEventHandlers();
156    }
157
158    protected SymbolicDataAnalysisProblem(T problemData, U evaluator, V solutionCreator)
159      : base(evaluator, solutionCreator) {
160      Parameters.Add(new ValueParameter<T>(ProblemDataParameterName, ProblemDataParameterDescription, problemData));
161      Parameters.Add(new ValueParameter<ISymbolicDataAnalysisGrammar>(SymbolicExpressionTreeGrammarParameterName, SymbolicExpressionTreeGrammarParameterDescription));
162      Parameters.Add(new ValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>(SymbolicExpressionTreeInterpreterParameterName, SymoblicExpressionTreeInterpreterParameterDescription));
163      Parameters.Add(new FixedValueParameter<IntValue>(MaximumSymbolicExpressionTreeDepthParameterName, MaximumSymbolicExpressionTreeDepthParameterDescription));
164      Parameters.Add(new FixedValueParameter<IntValue>(MaximumSymbolicExpressionTreeLengthParameterName, MaximumSymbolicExpressionTreeLengthParameterDescription));
165      Parameters.Add(new FixedValueParameter<IntValue>(MaximumFunctionDefinitionsParameterName, MaximumFunctionDefinitionsParameterDescription));
166      Parameters.Add(new FixedValueParameter<IntValue>(MaximumFunctionArgumentsParameterName, MaximumFunctionArgumentsParameterDescription));
167      Parameters.Add(new FixedValueParameter<IntRange>(FitnessCalculationPartitionParameterName, FitnessCalculationPartitionParameterDescription));
168      Parameters.Add(new FixedValueParameter<IntRange>(ValidationPartitionParameterName, ValidationPartitionParameterDescription));
169      Parameters.Add(new FixedValueParameter<PercentValue>(RelativeNumberOfEvaluatedSamplesParameterName, RelativeNumberOfEvaluatedSamplesParameterDescription, new PercentValue(1)));
170
171      SymbolicExpressionTreeInterpreterParameter.Hidden = true;
172      MaximumFunctionArgumentsParameter.Hidden = true;
173      MaximumFunctionDefinitionsParameter.Hidden = true;
174
175      SymbolicExpressionTreeGrammar = new TypeCoherentExpressionGrammar();
176      SymbolicExpressionTreeInterpreter = new SymbolicDataAnalysisExpressionTreeInterpreter();
177
178      FitnessCalculationPartition.Start = ProblemData.TrainingPartition.Start;
179      FitnessCalculationPartition.End = ProblemData.TrainingPartition.End;
180
181      InitializeOperators();
182
183      UpdateGrammar();
184      RegisterEventHandlers();
185    }
186
187    protected virtual void UpdateGrammar() {
188      SymbolicExpressionTreeGrammar.MaximumFunctionArguments = MaximumFunctionArguments.Value;
189      SymbolicExpressionTreeGrammar.MaximumFunctionDefinitions = MaximumFunctionDefinitions.Value;
190      foreach (var varSymbol in SymbolicExpressionTreeGrammar.Symbols.OfType<HeuristicLab.Problems.DataAnalysis.Symbolic.Variable>()) {
191        if (!varSymbol.Fixed) varSymbol.VariableNames = ProblemData.AllowedInputVariables;
192      }
193      foreach (var varSymbol in SymbolicExpressionTreeGrammar.Symbols.OfType<HeuristicLab.Problems.DataAnalysis.Symbolic.VariableCondition>()) {
194        if (!varSymbol.Fixed) varSymbol.VariableNames = ProblemData.AllowedInputVariables;
195      }
196    }
197
198    private void InitializeOperators() {
199      Operators.AddRange(ApplicationManager.Manager.GetInstances<ISymbolicExpressionTreeOperator>());
200      Operators.AddRange(ApplicationManager.Manager.GetInstances<ISymbolicDataAnalysisExpressionCrossover<T>>());
201      Operators.Add(new SymbolicExpressionSymbolFrequencyAnalyzer());
202      Operators.Add(new SymbolicDataAnalysisVariableFrequencyAnalyzer());
203      Operators.Add(new SymbolicDataAnalysisComplexityAnalyzer());
204      Operators.Add(new MinAverageMaxSymbolicExpressionTreeLengthAnalyzer());
205      Operators.Add(new SymbolicExpressionTreeLengthAnalyzer());
206      ParameterizeOperators();
207    }
208
209    #region events
210    private void RegisterEventHandlers() {
211      ProblemDataParameter.ValueChanged += new EventHandler(ProblemDataParameter_ValueChanged);
212      ProblemDataParameter.Value.Changed += (object sender, EventArgs e) => OnProblemDataChanged();
213
214      SymbolicExpressionTreeGrammarParameter.ValueChanged += new EventHandler(SymbolicExpressionTreeGrammarParameter_ValueChanged);
215
216      MaximumFunctionArguments.ValueChanged += new EventHandler(ArchitectureParameterValue_ValueChanged);
217      MaximumFunctionDefinitions.ValueChanged += new EventHandler(ArchitectureParameterValue_ValueChanged);
218      MaximumSymbolicExpressionTreeDepth.ValueChanged += new EventHandler(MaximumSymbolicExpressionTreeDepth_ValueChanged);
219    }
220
221    private void ProblemDataParameter_ValueChanged(object sender, EventArgs e) {
222      ValidationPartition.Start = 0;
223      ValidationPartition.End = 0;
224      ProblemDataParameter.Value.Changed += (object s, EventArgs args) => OnProblemDataChanged();
225      OnProblemDataChanged();
226    }
227
228    private void SymbolicExpressionTreeGrammarParameter_ValueChanged(object sender, EventArgs e) {
229      UpdateGrammar();
230    }
231
232    private void ArchitectureParameterValue_ValueChanged(object sender, EventArgs e) {
233      UpdateGrammar();
234    }
235
236    private void MaximumSymbolicExpressionTreeDepth_ValueChanged(object sender, EventArgs e) {
237      if (MaximumSymbolicExpressionTreeDepth != null && MaximumSymbolicExpressionTreeDepth.Value < 3)
238        MaximumSymbolicExpressionTreeDepth.Value = 3;
239    }
240
241    protected override void OnSolutionCreatorChanged() {
242      base.OnSolutionCreatorChanged();
243      SolutionCreator.SymbolicExpressionTreeParameter.ActualNameChanged += new EventHandler(SolutionCreator_SymbolicExpressionTreeParameter_ActualNameChanged);
244      ParameterizeOperators();
245    }
246
247    private void SolutionCreator_SymbolicExpressionTreeParameter_ActualNameChanged(object sender, EventArgs e) {
248      ParameterizeOperators();
249    }
250
251    protected override void OnEvaluatorChanged() {
252      base.OnEvaluatorChanged();
253      ParameterizeOperators();
254    }
255
256    public event EventHandler ProblemDataChanged;
257    protected virtual void OnProblemDataChanged() {
258      FitnessCalculationPartition.Start = ProblemData.TrainingPartition.Start;
259      FitnessCalculationPartition.End = ProblemData.TrainingPartition.End;
260
261      UpdateGrammar();
262      ParameterizeOperators();
263
264      var handler = ProblemDataChanged;
265      if (handler != null) handler(this, EventArgs.Empty);
266
267      OnReset();
268    }
269    #endregion
270
271    protected virtual void ParameterizeOperators() {
272      var operators = Parameters.OfType<IValueParameter>().Select(p => p.Value).OfType<IOperator>().Union(Operators).ToList();
273
274      foreach (var op in operators.OfType<ISymbolicExpressionTreeGrammarBasedOperator>()) {
275        op.SymbolicExpressionTreeGrammarParameter.ActualName = SymbolicExpressionTreeGrammarParameterName;
276      }
277      foreach (var op in operators.OfType<ISymbolicExpressionTreeSizeConstraintOperator>()) {
278        op.MaximumSymbolicExpressionTreeDepthParameter.ActualName = MaximumSymbolicExpressionTreeDepthParameterName;
279        op.MaximumSymbolicExpressionTreeLengthParameter.ActualName = MaximumSymbolicExpressionTreeLengthParameterName;
280      }
281      foreach (var op in operators.OfType<ISymbolicExpressionTreeArchitectureAlteringOperator>()) {
282        op.MaximumFunctionArgumentsParameter.ActualName = MaximumFunctionArgumentsParameterName;
283        op.MaximumFunctionDefinitionsParameter.ActualName = MaximumFunctionDefinitionsParameterName;
284      }
285      foreach (var op in operators.OfType<ISymbolicDataAnalysisEvaluator<T>>()) {
286        op.ProblemDataParameter.ActualName = ProblemDataParameterName;
287        op.SymbolicExpressionTreeParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
288        op.EvaluationPartitionParameter.ActualName = FitnessCalculationPartitionParameter.Name;
289        op.RelativeNumberOfEvaluatedSamplesParameter.ActualName = RelativeNumberOfEvaluatedSamplesParameter.Name;
290      }
291      foreach (var op in operators.OfType<ISymbolicExpressionTreeCrossover>()) {
292        op.ParentsParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
293        op.ChildParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
294      }
295      foreach (var op in operators.OfType<ISymbolicExpressionTreeManipulator>()) {
296        op.SymbolicExpressionTreeParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
297      }
298      foreach (var op in operators.OfType<ISymbolicExpressionTreeAnalyzer>()) {
299        op.SymbolicExpressionTreeParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
300      }
301      foreach (var op in operators.OfType<ISymbolicDataAnalysisAnalyzer>()) {
302        op.SymbolicExpressionTreeParameter.ActualName = SolutionCreator.SymbolicExpressionTreeParameter.ActualName;
303      }
304      foreach (var op in operators.OfType<ISymbolicDataAnalysisValidationAnalyzer<U, T>>()) {
305        op.RelativeNumberOfEvaluatedSamplesParameter.ActualName = RelativeNumberOfEvaluatedSamplesParameter.Name;
306        op.ValidationPartitionParameter.ActualName = ValidationPartitionParameter.Name;
307      }
308      foreach (var op in operators.OfType<ISymbolicDataAnalysisInterpreterOperator>()) {
309        op.SymbolicDataAnalysisTreeInterpreterParameter.ActualName = SymbolicExpressionTreeInterpreterParameterName;
310      }
311      foreach (var op in operators.OfType<ISymbolicDataAnalysisExpressionCrossover<T>>()) {
312        op.EvaluationPartitionParameter.ActualName = FitnessCalculationPartitionParameterName;
313      }
314      foreach (var op in operators.OfType<ISymbolicDataAnalysisExpressionCrossover<T>>()) {
315        op.ProblemDataParameter.ActualName = ProblemDataParameter.Name;
316        op.EvaluationPartitionParameter.ActualName = FitnessCalculationPartitionParameter.Name;
317        op.RelativeNumberOfEvaluatedSamplesParameter.ActualName = RelativeNumberOfEvaluatedSamplesParameter.Name;
318        op.EvaluatorParameter.ActualName = EvaluatorParameter.Name;
319      }
320    }
321
322    public abstract void ImportProblemDataFromFile(string fileName);
323  }
324}
Note: See TracBrowser for help on using the repository browser.