source: branches/HeuristicLab.OSGAEvaluator/HeuristicLab.OSGAEvaluator/SymbolicRegressionSingleObjectiveOSGAEvaluator.cs @ 14072

Last change on this file since 14072 was 14072, checked in by bburlacu, 5 years ago

#2635: Initial implementation.

File size: 15.7 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2016 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.Analysis;
26using HeuristicLab.Common;
27using HeuristicLab.Core;
28using HeuristicLab.Data;
29using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
30using HeuristicLab.Optimization;
31using HeuristicLab.Parameters;
32using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
33
34namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
35  [Item("SymbolicRegressionSingleObjectiveOSGAEvaluator", "An evaluator which tries to predict when a child will not be able to fullfil offspring selection criteria, to save evaluation time.")]
36  [StorableClass]
37  public class SymbolicRegressionSingleObjectiveOsgaEvaluator : SymbolicRegressionSingleObjectiveEvaluator {
38    private const string RelativeParentChildQualityThresholdParameterName = "RelativeParentChildQualityThreshold";
39    private const string RelativeFitnessEvaluationIntervalSizeParameterName = "RelativeFitnessEvaluationIntervalSize";
40    private const string ResultCollectionParameterName = "Results";
41
42    #region parameters
43    public ILookupParameter<ResultCollection> ResultCollectionParameter {
44      get { return (ILookupParameter<ResultCollection>)Parameters[ResultCollectionParameterName]; }
45    }
46    public IFixedValueParameter<IntValue> CorrectlyRejectedParameter {
47      get { return (IFixedValueParameter<IntValue>)Parameters["CorrectlyRejected"]; }
48    }
49    public IFixedValueParameter<IntValue> IncorrectlyRejectedParameter {
50      get { return (IFixedValueParameter<IntValue>)Parameters["IncorrectlyRejected"]; }
51    }
52    public IFixedValueParameter<IntValue> CorrectlyNotRejectedParameter {
53      get { return (IFixedValueParameter<IntValue>)Parameters["CorrectlyNotRejected"]; }
54    }
55    public IFixedValueParameter<IntValue> IncorrectlyNotRejectedParameter {
56      get { return (IFixedValueParameter<IntValue>)Parameters["IncorrectlyNotRejected"]; }
57    }
58    public IValueLookupParameter<DoubleValue> ComparisonFactorParameter {
59      get { return (ValueLookupParameter<DoubleValue>)Parameters["ComparisonFactor"]; }
60    }
61    public IFixedValueParameter<PercentValue> RelativeParentChildQualityThresholdParameter {
62      get { return (IFixedValueParameter<PercentValue>)Parameters[RelativeParentChildQualityThresholdParameterName]; }
63    }
64    public IFixedValueParameter<PercentValue> RelativeFitnessEvaluationIntervalSizeParameter {
65      get { return (IFixedValueParameter<PercentValue>)Parameters[RelativeFitnessEvaluationIntervalSizeParameterName]; }
66    }
67    public IScopeTreeLookupParameter<DoubleValue> ParentQualitiesParameter { get { return (IScopeTreeLookupParameter<DoubleValue>)Parameters["ParentQualities"]; } }
68    #endregion
69
70    #region parameter properties
71    public double RelativeParentChildQualityThreshold {
72      get { return RelativeParentChildQualityThresholdParameter.Value.Value; }
73      set { RelativeParentChildQualityThresholdParameter.Value.Value = value; }
74    }
75
76    public double RelativeFitnessEvaluationIntervalSize {
77      get { return RelativeFitnessEvaluationIntervalSizeParameter.Value.Value; }
78      set { RelativeFitnessEvaluationIntervalSizeParameter.Value.Value = value; }
79    }
80
81    public int CorrectlyRejected {
82      get { return CorrectlyRejectedParameter.Value.Value; }
83      set { CorrectlyRejectedParameter.Value.Value = value; }
84    }
85
86    public int CorrectlyNotRejected {
87      get { return CorrectlyNotRejectedParameter.Value.Value; }
88      set { CorrectlyNotRejectedParameter.Value.Value = value; }
89    }
90
91    public int IncorrectlyRejected {
92      get { return IncorrectlyRejectedParameter.Value.Value; }
93      set { IncorrectlyRejectedParameter.Value.Value = value; }
94    }
95
96    public int IncorrectlyNotRejected {
97      get { return IncorrectlyNotRejectedParameter.Value.Value; }
98      set { IncorrectlyNotRejectedParameter.Value.Value = value; }
99    }
100    #endregion
101
102    public override bool Maximization {
103      get { return true; }
104    }
105
106    public SymbolicRegressionSingleObjectiveOsgaEvaluator() {
107      Parameters.Add(new ValueLookupParameter<DoubleValue>("ComparisonFactor", "Determines if the quality should be compared to the better parent (1.0), to the worse (0.0) or to any linearly interpolated value between them."));
108      Parameters.Add(new FixedValueParameter<PercentValue>(RelativeParentChildQualityThresholdParameterName, new PercentValue(0.1)));
109      Parameters.Add(new FixedValueParameter<PercentValue>(RelativeFitnessEvaluationIntervalSizeParameterName, new PercentValue(0.1)));
110      Parameters.Add(new FixedValueParameter<IntValue>("CorrectlyRejected", new IntValue(0)));
111      Parameters.Add(new FixedValueParameter<IntValue>("IncorrectlyRejected", new IntValue(0)));
112      Parameters.Add(new FixedValueParameter<IntValue>("CorrectlyNotRejected", new IntValue(0)));
113      Parameters.Add(new FixedValueParameter<IntValue>("IncorrectlyNotRejected", new IntValue(0)));
114      Parameters.Add(new LookupParameter<ResultCollection>(ResultCollectionParameterName));
115      Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("ParentQualities") { ActualName = "Quality" });
116    }
117
118    [StorableHook(HookType.AfterDeserialization)]
119    private void AfterDeserialization() {
120      if (!Parameters.ContainsKey(ResultCollectionParameterName))
121        Parameters.Add(new LookupParameter<ResultCollection>(ResultCollectionParameterName));
122
123      if (!Parameters.ContainsKey("ParentQualities"))
124        Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("ParentQualities") { ActualName = "Quality" });
125    }
126
127    [StorableConstructor]
128    protected SymbolicRegressionSingleObjectiveOsgaEvaluator(bool deserializing) : base(deserializing) { }
129
130    protected SymbolicRegressionSingleObjectiveOsgaEvaluator(SymbolicRegressionSingleObjectiveOsgaEvaluator original, Cloner cloner) : base(original, cloner) { }
131
132    public override IDeepCloneable Clone(Cloner cloner) {
133      return new SymbolicRegressionSingleObjectiveOsgaEvaluator(this, cloner);
134    }
135
136    public override void ClearState() {
137      base.ClearState();
138      CorrectlyNotRejected = 0;
139      CorrectlyRejected = 0;
140      IncorrectlyNotRejected = 0;
141      IncorrectlyRejected = 0;
142    }
143
144    public override IOperation InstrumentedApply() {
145      var solution = SymbolicExpressionTreeParameter.ActualValue;
146      IEnumerable<int> rows = GenerateRowsToEvaluate();
147
148      var interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
149      var estimationLimits = EstimationLimitsParameter.ActualValue;
150      var problemData = ProblemDataParameter.ActualValue;
151      var applyLinearScaling = ApplyLinearScalingParameter.ActualValue.Value;
152
153      double quality;
154      var parentQualities = ParentQualitiesParameter.ActualValue;
155
156      // parent subscopes are not present during evaluation of the initial population
157      if (parentQualities.Length > 0) {
158        quality = Calculate(interpreter, solution, estimationLimits, problemData, rows, applyLinearScaling);
159      } else {
160        quality = Calculate(interpreter, solution, estimationLimits.Lower, estimationLimits.Upper, problemData, rows, applyLinearScaling);
161      }
162      QualityParameter.ActualValue = new DoubleValue(quality);
163
164      return base.InstrumentedApply();
165    }
166
167    public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
168      IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
169      IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
170      OnlineCalculatorError errorState;
171
172      double r;
173      if (applyLinearScaling) {
174        var rCalculator = new OnlinePearsonsRCalculator();
175        CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, rCalculator, problemData.Dataset.Rows);
176        errorState = rCalculator.ErrorState;
177        r = rCalculator.R;
178      } else {
179        IEnumerable<double> boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
180        r = OnlinePearsonsRCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
181      }
182      if (errorState != OnlineCalculatorError.None) return double.NaN;
183      return r * r;
184    }
185
186    private double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, DoubleLimit estimationLimits, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
187      var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows).ToList();
188      var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
189      IEnumerator<double> targetValuesEnumerator;
190
191      double alpha = 0, beta = 1;
192      if (applyLinearScaling) {
193        var linearScalingCalculator = new OnlineLinearScalingParameterCalculator();
194        targetValuesEnumerator = targetValues.GetEnumerator();
195        var estimatedValuesEnumerator = estimatedValues.GetEnumerator();
196        while (targetValuesEnumerator.MoveNext() & estimatedValuesEnumerator.MoveNext()) {
197          double target = targetValuesEnumerator.Current;
198          double estimated = estimatedValuesEnumerator.Current;
199          if (!double.IsNaN(estimated) && !double.IsInfinity(estimated))
200            linearScalingCalculator.Add(estimated, target);
201        }
202        if (linearScalingCalculator.ErrorState == OnlineCalculatorError.None && (targetValuesEnumerator.MoveNext() || estimatedValuesEnumerator.MoveNext()))
203          throw new ArgumentException("Number of elements in target and estimated values enumeration do not match.");
204
205        alpha = linearScalingCalculator.Alpha;
206        beta = linearScalingCalculator.Beta;
207        if (linearScalingCalculator.ErrorState != OnlineCalculatorError.None) {
208          alpha = 0.0;
209          beta = 1.0;
210        }
211      }
212      var scaledEstimatedValuesEnumerator = estimatedValues.Select(x => x * beta + alpha).LimitToRange(estimationLimits.Lower, estimationLimits.Upper).GetEnumerator();
213      targetValuesEnumerator = targetValues.GetEnumerator();
214
215      var pearsonRCalculator = new OnlinePearsonsRCalculator();
216
217      var interval = (int)Math.Floor(problemData.TrainingPartition.Size * RelativeFitnessEvaluationIntervalSize);
218      var i = 0;
219      var qualityPerInterval = new List<double>();
220      while (targetValuesEnumerator.MoveNext() && scaledEstimatedValuesEnumerator.MoveNext()) {
221        pearsonRCalculator.Add(targetValuesEnumerator.Current, scaledEstimatedValuesEnumerator.Current);
222        ++i;
223        if (i % interval == 0) {
224          var q = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
225          qualityPerInterval.Add(q * q);
226        }
227      }
228      var r = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
229      var actualQuality = r * r;
230      var parentQualities = ParentQualitiesParameter.ActualValue.Select(x => x.Value);
231      var minQuality = parentQualities.Min();
232      var maxQuality = parentQualities.Max();
233      var comparisonFactor = ComparisonFactorParameter.ActualValue.Value;
234      var parentQuality = minQuality + (maxQuality - minQuality) * comparisonFactor;
235      var threshold = parentQuality * RelativeParentChildQualityThreshold;
236
237      //var predictedRejected = qualityPerInterval.Any(x => double.IsNaN(x) || !(x > threshold));
238
239      bool predictedRejected = false;
240
241      DataTable table;
242      var results = ResultCollectionParameter.ActualValue;
243      if (!results.ContainsKey("RejectionCounts")) {
244        table = new DataTable("RejectionCounts");
245        results.Add(new Result("RejectionCounts", table));
246
247        var row = new DataRow("Predicted Rejected") { VisualProperties = { ChartType = DataRowVisualProperties.DataRowChartType.Histogram } };
248        table.Rows.Add(row);
249
250        row = new DataRow("Actually Rejected") { VisualProperties = { ChartType = DataRowVisualProperties.DataRowChartType.Histogram } };
251        table.Rows.Add(row);
252
253        //        row = new DataRow("Actually Not Rejected") { VisualProperties = { ChartType = DataRowVisualProperties.DataRowChartType.Columns, StartIndexZero = true } };
254        //        row.Values.AddRange(qualityPerInterval.Select(x => 0.0));
255        //        table.Rows.Add(row);
256        //
257        //        row = new DataRow("Predicted Not Rejected") { VisualProperties = { ChartType = DataRowVisualProperties.DataRowChartType.Columns, StartIndexZero = true } };
258        //        row.Values.AddRange(qualityPerInterval.Select(x => 0.0));
259        //        table.Rows.Add(row);
260      } else {
261        table = (DataTable)results["RejectionCounts"].Value;
262      }
263
264      i = 0;
265      foreach (var q in qualityPerInterval) {
266        if (double.IsNaN(q) || !(q > threshold)) {
267          predictedRejected = true;
268          break;
269        }
270        ++i;
271      }
272
273      var actuallyRejected = !(actualQuality > parentQuality);
274      if (predictedRejected) {
275        table.Rows["Predicted Rejected"].Values.Add(i);
276        if (actuallyRejected)
277          table.Rows["Actually Rejected"].Values.Add(i);
278      }
279      //      else {
280      //        table.Rows["Predicted Not Rejected"].Values[i]++;
281      //        if (!actuallyRejected)
282      //          table.Rows["Actually Not Rejected"].Values[i]++;
283      //      }
284
285      if (predictedRejected) {
286        if (actuallyRejected) {
287          CorrectlyRejected++;
288        } else {
289          IncorrectlyRejected++;
290        }
291      } else {
292        if (actuallyRejected) {
293          IncorrectlyNotRejected++;
294        } else {
295          CorrectlyNotRejected++;
296        }
297      }
298      return r * r;
299    }
300
301    public override double Evaluate(IExecutionContext context, ISymbolicExpressionTree tree, IRegressionProblemData problemData, IEnumerable<int> rows) {
302      SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = context;
303      EstimationLimitsParameter.ExecutionContext = context;
304      ApplyLinearScalingParameter.ExecutionContext = context;
305
306      var interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
307      var estimationLimits = EstimationLimitsParameter.ActualValue;
308      var applyLinearScaling = ApplyLinearScalingParameter.ActualValue.Value;
309
310      double r2 = Calculate(interpreter, tree, estimationLimits.Lower, estimationLimits.Upper, problemData, rows, applyLinearScaling);
311
312      SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = null;
313      EstimationLimitsParameter.ExecutionContext = null;
314      ApplyLinearScalingParameter.ExecutionContext = null;
315
316      return r2;
317    }
318  }
319}
Note: See TracBrowser for help on using the repository browser.