source: branches/2974_Constants_Optimization/UnitTests/ConstantsOptimizationTests.cs @ 18183

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2018 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
 *
 * This file is part of HeuristicLab.
 *
 * HeuristicLab is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * HeuristicLab is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
 */
#endregion

using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using HeuristicLab.Problems.DataAnalysis;
using HeuristicLab.Problems.DataAnalysis.Symbolic;
using HeuristicLab.Problems.DataAnalysis.Symbolic.ConstantsOptimization;
using HeuristicLab.Problems.DataAnalysis.Symbolic.Regression;


using HeuristicLab.Problems.Instances.DataAnalysis;
using Microsoft.VisualStudio.TestTools.UnitTesting;

namespace UnitTests {
  [TestClass]
  public class ConstantsOptimizationTests {
    private static readonly int seed = 1234;


    public static void CompareConstantsOptimizationResults(IRegressionProblemData problemData, ISymbolicExpressionTree tree) {
      var applyLinearScaling = true;
      var old_optimizedTree = (ISymbolicExpressionTree)tree.Clone();
      var old_result = SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(
        new SymbolicDataAnalysisExpressionTreeLinearInterpreter(),
        old_optimizedTree, problemData, problemData.TrainingIndices, applyLinearScaling, maxIterations: 10);


      var new_optimizedTree = (ISymbolicExpressionTree)tree.Clone();
      var new_result = LMConstantsOptimizer.OptimizeConstants(new_optimizedTree, problemData.Dataset, problemData.TargetVariable, problemData.TrainingIndices, applyLinearScaling, maxIterations: 10);

      // extract constants
      var old_constants = Util.ExtractConstants(old_optimizedTree, applyLinearScaling);
      var new_constants = Util.ExtractConstants(new_optimizedTree, applyLinearScaling);

      {
        // consistency with old ConstOpt style
        var initalR2 = SymbolicRegressionSingleObjectivePearsonRSquaredEvaluator.Calculate(
          new SymbolicDataAnalysisExpressionTreeLinearInterpreter(),
          tree, double.MinValue, double.MaxValue, problemData, problemData.TrainingIndices, applyLinearScaling);

        // LMConstantsOptimizer returns 0 if optimization was not successful
        if (initalR2 - new_result > 0.001) {
          new_result = initalR2;
          new_constants = old_constants;
        }
      }

      //check R² values
      Assert.AreEqual(old_result, new_result, 1E-6);

      //check numeric values of constants
      for (int i = 0; i < old_constants.Length; i++) {
        if (old_constants[i] == 0) {
          Assert.AreEqual(old_constants[i], new_constants[i], 1E-8); // check absolute error
        } else {
          Assert.AreEqual(0.0, 1.0 - new_constants[i] / old_constants[i], 1E-6); // check percentual error
        }
      }
    }



    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Nguyen01() {
      var problemData = new NguyenFunctionOne(seed).GenerateRegressionData();
      var modelTemplate = "({0})* CUBE(X) + ({1}) * SQR(X) + ({2}) * X";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, -2.0, 500.362 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { -6987.25, 1, -888 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Nguyen03() {
      var problemData = new NguyenFunctionThree(seed).GenerateRegressionData();
      var modelTemplate = "({0})* X*X*X*X*X + ({1}) * X*X*X*X + ({2}) * X*X*X + ({3}) * X*X + ({4}) * X + {5}";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0, 4.0, 5.0, 6.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, -2.0, 500.362, -5646, 0.0001, 1234 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { -6987.25, 1, -888, +888, -1, 6987.25, 0, 25 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Nguyen05() {
      var problemData = new NguyenFunctionFive(seed).GenerateRegressionData();
      var modelTemplate = "({0}) * SIN(({1})*X*X) * COS(({2})*X) + ({3})";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0, 4.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, -2.0, -3.0, 5.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 0.5, 1, 1, 3 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Nguyen07() {
      var problemData = new NguyenFunctionSeven(seed).GenerateRegressionData();
      var modelTemplate = "({0}) * LOG(({1})*X + ({2})) + ({3}) * LOG(({4})*X*X + ({5})) + ({6})";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, 2.0, 500.362, -5646, 0.0001, 1234, 421 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      // This test fails not because of a problem in ConstOpt but because of the code in 
      // OnlinePearsonsRCalculator
      //         double xVar = sxCalculator.PopulationVariance;
      //         double yVar = syCalculator.PopulationVariance;
      //         if (xVar.IsAlmost(0.0) || yVar.IsAlmost(0.0)) {
      //           return 0.0;
      //         } else { ...
      // PopulationVariance might become close to zero, but the result of cov / (sqrt(xvar) * sqrt(yvar) might still be correct.
      // Currently, we just return 0. Which means that we reset optimized constants to initial values
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { -6987.25, 1, 888, +888, -1, 6987.25, 0, 25 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Nguyen08() {
      var problemData = new NguyenFunctionEight(seed).GenerateRegressionData();
      var modelTemplate = "({0})* SQRT(({1}) * X) + ({2})";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, 20.0, -500.362 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { -6987.25, 1, -888 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestMethod]
    [TestCategory("Problems.DataAnalysis.Symbolic.Regression")]
    [TestProperty("Time", "short")]
    public void ConstantsOptimizationTest_Keijzer05() {
      var problemData = new KeijzerFunctionFive(seed).GenerateRegressionData();
      var modelTemplate = "( ({0}) * X * Z ) / ( (({1}) * X + ({2})) * ({3}) * SQR(Y) ) + ({4})";

      string modelString;
      object[] constants;
      ISymbolicExpressionTree modelTree;

      constants = new object[] { 1.0, 2.0, 3.0, 4.0, 5.0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { 5.0, -2.0, 500.362, -5646, 0.0001 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);

      constants = new object[] { -6987.25, 1, -888, +888, -1, 6987.25, 0 };
      modelString = string.Format(modelTemplate, constants);
      modelTree = new InfixExpressionParser().Parse(modelString);
      CompareConstantsOptimizationResults(problemData, modelTree);
    }

    [TestInitialize]
    public void InitEnvironment() {
      System.Threading.Thread.CurrentThread.CurrentCulture = System.Globalization.CultureInfo.InvariantCulture;
    }
  }
}