Free cookie consent management tool by TermsFeed Policy Generator

source: branches/ALPS/HeuristicLab.Tests/HeuristicLab-3.3/Samples/GPSymbolicRegressionSampleTest.cs @ 11655

Last change on this file since 11655 was 11450, checked in by bburlacu, 10 years ago

#2211: Separated samples class into separate test classes. Added scripts unit tests (grid search classification/regression).

File size: 8.3 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2014 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.IO;
23using System.Linq;
24using HeuristicLab.Algorithms.GeneticAlgorithm;
25using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
26using HeuristicLab.Persistence.Default.Xml;
27using HeuristicLab.Problems.DataAnalysis;
28using HeuristicLab.Problems.DataAnalysis.Symbolic;
29using HeuristicLab.Problems.DataAnalysis.Symbolic.Regression;
30using HeuristicLab.Problems.Instances.DataAnalysis;
31using HeuristicLab.Selection;
32using Microsoft.VisualStudio.TestTools.UnitTesting;
33
34namespace HeuristicLab.Tests {
35  /// <summary>
36  /// Summary description for GPSymbolicRegressionSampleTest
37  /// </summary>
38  [TestClass]
39  public class GPSymbolicRegressionSampleTest {
40    private const string samplesDirectory = SamplesUtils.Directory;
41    [ClassInitialize]
42    public static void MyClassInitialize(TestContext testContext) {
43      if (!Directory.Exists(samplesDirectory))
44        Directory.CreateDirectory(samplesDirectory);
45    }
46
47
48    [TestMethod]
49    [TestCategory("Samples.Create")]
50    [TestProperty("Time", "medium")]
51    public void CreateGpSymbolicRegressionSampleTest() {
52      var ga = CreateGpSymbolicRegressionSample();
53      XmlGenerator.Serialize(ga, @"Samples\SGP_SymbReg.hl");
54    }
55    [TestMethod]
56    [TestCategory("Samples.Execute")]
57    [TestProperty("Time", "long")]
58    public void RunGpSymbolicRegressionSampleTest() {
59      var ga = CreateGpSymbolicRegressionSample();
60      ga.SetSeedRandomly.Value = false;
61      SamplesUtils.RunAlgorithm(ga);
62      Assert.AreEqual(0.858344291534625, SamplesUtils.GetDoubleResult(ga, "BestQuality"), 1E-8);
63      Assert.AreEqual(0.56758466520692641, SamplesUtils.GetDoubleResult(ga, "CurrentAverageQuality"), 1E-8);
64      Assert.AreEqual(0, SamplesUtils.GetDoubleResult(ga, "CurrentWorstQuality"), 1E-8);
65      Assert.AreEqual(50950, SamplesUtils.GetIntResult(ga, "EvaluatedSolutions"));
66      var bestTrainingSolution = (IRegressionSolution)ga.Results["Best training solution"].Value;
67      Assert.AreEqual(0.85504801557844745, bestTrainingSolution.TrainingRSquared, 1E-8);
68      Assert.AreEqual(0.86259381948647817, bestTrainingSolution.TestRSquared, 1E-8);
69      var bestValidationSolution = (IRegressionSolution)ga.Results["Best validation solution"].Value;
70      Assert.AreEqual(0.84854338315539746, bestValidationSolution.TrainingRSquared, 1E-8);
71      Assert.AreEqual(0.8662813452656678, bestValidationSolution.TestRSquared, 1E-8);
72    }
73
74    private GeneticAlgorithm CreateGpSymbolicRegressionSample() {
75      GeneticAlgorithm ga = new GeneticAlgorithm();
76      #region Problem Configuration
77      SymbolicRegressionSingleObjectiveProblem symbRegProblem = new SymbolicRegressionSingleObjectiveProblem();
78      symbRegProblem.Name = "Tower Symbolic Regression Problem";
79      symbRegProblem.Description = "Tower Dataset (downloaded from: http://www.symbolicregression.com/?q=towerProblem)";
80      RegressionRealWorldInstanceProvider provider = new RegressionRealWorldInstanceProvider();
81      var instance = provider.GetDataDescriptors().Where(x => x.Name.Equals("Tower")).Single();
82      var towerProblemData = (RegressionProblemData)provider.LoadData(instance);
83      towerProblemData.TargetVariableParameter.Value = towerProblemData.TargetVariableParameter.ValidValues
84        .First(v => v.Value == "towerResponse");
85      towerProblemData.InputVariables.SetItemCheckedState(
86        towerProblemData.InputVariables.Single(x => x.Value == "x1"), true);
87      towerProblemData.InputVariables.SetItemCheckedState(
88        towerProblemData.InputVariables.Single(x => x.Value == "x7"), false);
89      towerProblemData.InputVariables.SetItemCheckedState(
90        towerProblemData.InputVariables.Single(x => x.Value == "x11"), false);
91      towerProblemData.InputVariables.SetItemCheckedState(
92        towerProblemData.InputVariables.Single(x => x.Value == "x16"), false);
93      towerProblemData.InputVariables.SetItemCheckedState(
94        towerProblemData.InputVariables.Single(x => x.Value == "x21"), false);
95      towerProblemData.InputVariables.SetItemCheckedState(
96        towerProblemData.InputVariables.Single(x => x.Value == "x25"), false);
97      towerProblemData.InputVariables.SetItemCheckedState(
98        towerProblemData.InputVariables.Single(x => x.Value == "towerResponse"), false);
99      towerProblemData.TrainingPartition.Start = 0;
100      towerProblemData.TrainingPartition.End = 3136;
101      towerProblemData.TestPartition.Start = 3136;
102      towerProblemData.TestPartition.End = 4999;
103      towerProblemData.Name = "Data imported from towerData.txt";
104      towerProblemData.Description = "Chemical concentration at top of distillation tower, dataset downloaded from: http://vanillamodeling.com/realproblems.html, best R² achieved with nu-SVR = 0.97";
105      symbRegProblem.ProblemData = towerProblemData;
106
107      // configure grammar
108      var grammar = new TypeCoherentExpressionGrammar();
109      grammar.ConfigureAsDefaultRegressionGrammar();
110      grammar.Symbols.OfType<VariableCondition>().Single().InitialFrequency = 0.0;
111      var varSymbol = grammar.Symbols.OfType<Variable>().Where(x => !(x is LaggedVariable)).Single();
112      varSymbol.WeightMu = 1.0;
113      varSymbol.WeightSigma = 1.0;
114      varSymbol.WeightManipulatorMu = 0.0;
115      varSymbol.WeightManipulatorSigma = 0.05;
116      varSymbol.MultiplicativeWeightManipulatorSigma = 0.03;
117      var constSymbol = grammar.Symbols.OfType<Constant>().Single();
118      constSymbol.MaxValue = 20;
119      constSymbol.MinValue = -20;
120      constSymbol.ManipulatorMu = 0.0;
121      constSymbol.ManipulatorSigma = 1;
122      constSymbol.MultiplicativeManipulatorSigma = 0.03;
123      symbRegProblem.SymbolicExpressionTreeGrammar = grammar;
124
125      // configure remaining problem parameters
126      symbRegProblem.BestKnownQuality.Value = 0.97;
127      symbRegProblem.FitnessCalculationPartition.Start = 0;
128      symbRegProblem.FitnessCalculationPartition.End = 2300;
129      symbRegProblem.ValidationPartition.Start = 2300;
130      symbRegProblem.ValidationPartition.End = 3136;
131      symbRegProblem.RelativeNumberOfEvaluatedSamples.Value = 1;
132      symbRegProblem.MaximumSymbolicExpressionTreeLength.Value = 150;
133      symbRegProblem.MaximumSymbolicExpressionTreeDepth.Value = 12;
134      symbRegProblem.MaximumFunctionDefinitions.Value = 0;
135      symbRegProblem.MaximumFunctionArguments.Value = 0;
136
137      symbRegProblem.EvaluatorParameter.Value = new SymbolicRegressionSingleObjectivePearsonRSquaredEvaluator();
138      #endregion
139      #region Algorithm Configuration
140      ga.Problem = symbRegProblem;
141      ga.Name = "Genetic Programming - Symbolic Regression";
142      ga.Description = "A standard genetic programming algorithm to solve a symbolic regression problem (tower dataset)";
143      SamplesUtils.ConfigureGeneticAlgorithmParameters<TournamentSelector, SubtreeCrossover, MultiSymbolicExpressionTreeManipulator>(
144        ga, 1000, 1, 50, 0.15, 5);
145      var mutator = (MultiSymbolicExpressionTreeManipulator)ga.Mutator;
146      mutator.Operators.OfType<FullTreeShaker>().Single().ShakingFactor = 0.1;
147      mutator.Operators.OfType<OnePointShaker>().Single().ShakingFactor = 1.0;
148
149      ga.Analyzer.Operators.SetItemCheckedState(
150        ga.Analyzer.Operators
151        .OfType<SymbolicRegressionSingleObjectiveOverfittingAnalyzer>()
152        .Single(), false);
153      ga.Analyzer.Operators.SetItemCheckedState(
154        ga.Analyzer.Operators
155        .OfType<SymbolicDataAnalysisAlleleFrequencyAnalyzer>()
156        .First(), false);
157      #endregion
158      return ga;
159    }
160  }
161}
Note: See TracBrowser for help on using the repository browser.