source: branches/DataAnalysis/HeuristicLab.Problems.DataAnalysis.MultiVariate.Regression/3.3/Symbolic/Evaluators/PartialDerivativeEvaluator.cs @ 10764

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2008 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.Core;
using HeuristicLab.Data;
using System;
using HeuristicLab.Common;
using System.Linq;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
using HeuristicLab.Problems.DataAnalysis.Symbolic;
using System.Collections.Generic;
using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using HeuristicLab.Problems.DataAnalysis.Regression.Symbolic;
using HeuristicLab.Problems.DataAnalysis.MultiVariate.Regression.Symbolic.Interfaces;
using HeuristicLab.Parameters;


namespace HeuristicLab.Problems.DataAnalysis.MultiVariate.Regression.Symbolic.Evaluators {
  [Item("PartialDerivativeEvaluator", "Evaluator for implict equation modelling")]
  [StorableClass]
  public class PartialDerivativeEvaluator : SingleObjectiveSymbolicVectorRegressionEvaluator {

    [StorableConstructor]
    protected PartialDerivativeEvaluator(bool deserializing) : base(deserializing) { }
    protected PartialDerivativeEvaluator(PartialDerivativeEvaluator original, Cloner cloner)
      : base(original, cloner) {
    }
    public PartialDerivativeEvaluator()
      : base() {
    }
    public override IDeepCloneable Clone(Cloner cloner) {
      return new PartialDerivativeEvaluator(this, cloner);
    }
    public override double Evaluate(SymbolicExpressionTree tree, ISymbolicExpressionTreeInterpreter interpreter, MultiVariateDataAnalysisProblemData problemData, IEnumerable<string> targetVariables, IEnumerable<int> rows, DoubleArray lowerEstimationBound, DoubleArray upperEstimationBound) {

      Dataset dataset = problemData.Dataset;
      IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(tree, dataset, rows);

      var sortedNames = targetVariables.OrderBy(x => x);
      var pairs = from v1 in sortedNames
                  from v2 in sortedNames.Skip(1)
                  where v1.CompareTo(v2) < 0
                  select new { x = v1, y = v2 };

      double meanErrorSum = 0;
      foreach (var pair in pairs) {
        double errorSum = 0;
        string variableX = pair.x;
        string variableY = pair.y;

        var dFdX = new SymbolicExpressionTree(PartialSymbolicDifferential.Apply((SymbolicExpressionTreeNode)tree.Root.Clone(), variableX, variableY));
        IEnumerable<double> estimatedDfDx = interpreter.GetSymbolicExpressionTreeValues(dFdX, dataset, rows).ToList();

        var dFdY = new SymbolicExpressionTree(PartialSymbolicDifferential.Apply((SymbolicExpressionTreeNode)tree.Root.Clone(), variableY, variableX));
        IEnumerable<double> estimatedDfDy = interpreter.GetSymbolicExpressionTreeValues(dFdY, dataset, rows).ToList();

        List<int> rowsList = rows.ToList();
        int n = rowsList.Count;
        int x = dataset.GetVariableIndex(variableX);
        int y = dataset.GetVariableIndex(variableY);

        var estimatedDfDxEnumerator = estimatedDfDx.GetEnumerator();
        var estimatedDfDyEnumerator = estimatedDfDy.GetEnumerator();
        var rowsEnumerator = rows.GetEnumerator();

        // skip 1
        estimatedDfDxEnumerator.MoveNext();
        estimatedDfDyEnumerator.MoveNext();
        rowsEnumerator.MoveNext();

        for (int i = 1; i < n - 1; i++) {
          // evaluate next
          estimatedDfDxEnumerator.MoveNext();
          estimatedDfDyEnumerator.MoveNext();
          rowsEnumerator.MoveNext();

          double dFdXValue = estimatedDfDxEnumerator.Current;
          double dFdYValue = estimatedDfDyEnumerator.Current;
          double dXdY = GetLocalDifferential(dataset, rowsEnumerator.Current, x, y);

          if ((dFdXValue.IsAlmost(0.0) && dFdYValue.IsAlmost(0.0))) {
            errorSum += Math.Log(1 + Math.Abs(dXdY));
          } else if (dFdXValue.IsAlmost(0.0) ||
            double.IsInfinity(dFdXValue) || double.IsNaN(dFdXValue) ||
            double.IsInfinity(dFdYValue) || double.IsNaN(dFdYValue)) {
            errorSum += 1000000;
          } else {
            double error = dXdY - dFdYValue / dFdXValue;
            errorSum += Math.Log(1 + Math.Abs(error));
            // errorSum += error * error;
          }
        }

        meanErrorSum += errorSum / n;
      }
      meanErrorSum /= pairs.Count();

      return meanErrorSum;
    }

    private double GetLocalDifferential(Dataset dataset, int i, int varX, int varY) {
      return
        (dataset[i + 1, varX] - dataset[i - 1, varX]) /
        (dataset[i + 1, varY] - dataset[i - 1, varY]);
    }

  }
}