source: branches/3128_Prediction_Intervals/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionSolution.cs @ 17991

#region License Information
/* HeuristicLab
 * Copyright (C) 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 System.Collections.Generic;
using System.Linq;
using HEAL.Attic;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
using HeuristicLab.Optimization;

namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
  /// <summary>
  /// Represents a symbolic regression solution (model + data) and attributes of the solution like accuracy and complexity
  /// </summary>
  [StorableType("88E56AF9-AD72-47E4-A613-8875703BD927")]
  [Item(Name = "SymbolicRegressionSolution", Description = "Represents a symbolic regression solution (model + data) and attributes of the solution like accuracy and complexity.")]
  public sealed class SymbolicRegressionSolution : RegressionSolution, ISymbolicRegressionSolution {
    private const string ModelLengthResultName = "Model Length";
    private const string ModelDepthResultName = "Model Depth";

    private const string EstimationLimitsResultsResultName = "Estimation Limits Results";
    private const string EstimationLimitsResultName = "Estimation Limits";
    private const string TrainingUpperEstimationLimitHitsResultName = "Training Upper Estimation Limit Hits";
    private const string TestLowerEstimationLimitHitsResultName = "Test Lower Estimation Limit Hits";
    private const string TrainingLowerEstimationLimitHitsResultName = "Training Lower Estimation Limit Hits";
    private const string TestUpperEstimationLimitHitsResultName = "Test Upper Estimation Limit Hits";
    private const string TrainingNaNEvaluationsResultName = "Training NaN Evaluations";
    private const string TestNaNEvaluationsResultName = "Test NaN Evaluations";

    private const string ModelBoundsResultName = "Model Bounds";

    public new ISymbolicRegressionModel Model {
      get { return (ISymbolicRegressionModel)base.Model; }
      set { base.Model = value; }
    }
    ISymbolicDataAnalysisModel ISymbolicDataAnalysisSolution.Model {
      get { return (ISymbolicDataAnalysisModel)base.Model; }
    }
    public int ModelLength {
      get { return ((IntValue)this[ModelLengthResultName].Value).Value; }
      private set { ((IntValue)this[ModelLengthResultName].Value).Value = value; }
    }

    public int ModelDepth {
      get { return ((IntValue)this[ModelDepthResultName].Value).Value; }
      private set { ((IntValue)this[ModelDepthResultName].Value).Value = value; }
    }

    private ResultCollection EstimationLimitsResultCollection {
      get { return (ResultCollection)this[EstimationLimitsResultsResultName].Value; }
    }
    public DoubleLimit EstimationLimits {
      get { return (DoubleLimit)EstimationLimitsResultCollection[EstimationLimitsResultName].Value; }
    }

    public int TrainingUpperEstimationLimitHits {
      get { return ((IntValue)EstimationLimitsResultCollection[TrainingUpperEstimationLimitHitsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TrainingUpperEstimationLimitHitsResultName].Value).Value = value; }
    }
    public int TestUpperEstimationLimitHits {
      get { return ((IntValue)EstimationLimitsResultCollection[TestUpperEstimationLimitHitsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TestUpperEstimationLimitHitsResultName].Value).Value = value; }
    }
    public int TrainingLowerEstimationLimitHits {
      get { return ((IntValue)EstimationLimitsResultCollection[TrainingLowerEstimationLimitHitsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TrainingLowerEstimationLimitHitsResultName].Value).Value = value; }
    }
    public int TestLowerEstimationLimitHits {
      get { return ((IntValue)EstimationLimitsResultCollection[TestLowerEstimationLimitHitsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TestLowerEstimationLimitHitsResultName].Value).Value = value; }
    }
    public int TrainingNaNEvaluations {
      get { return ((IntValue)EstimationLimitsResultCollection[TrainingNaNEvaluationsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TrainingNaNEvaluationsResultName].Value).Value = value; }
    }
    public int TestNaNEvaluations {
      get { return ((IntValue)EstimationLimitsResultCollection[TestNaNEvaluationsResultName].Value).Value; }
      private set { ((IntValue)EstimationLimitsResultCollection[TestNaNEvaluationsResultName].Value).Value = value; }
    }

    public IntervalCollection ModelBoundsCollection {
      get {
        if (!ContainsKey(ModelBoundsResultName)) return null;
        return (IntervalCollection)this[ModelBoundsResultName].Value;
      }
      private set {
        if (ContainsKey(ModelBoundsResultName)) {
          this[ModelBoundsResultName].Value = value;
        } else {
          Add(new Result(ModelBoundsResultName, "Results concerning the derivation of symbolic regression solution", value));
        }
      }
    }

    IConfidenceRegressionModel IConfidenceRegressionSolution.Model => Model;

    [StorableConstructor]
    private SymbolicRegressionSolution(StorableConstructorFlag _) : base(_) { }
    private SymbolicRegressionSolution(SymbolicRegressionSolution original, Cloner cloner)
      : base(original, cloner) {
    }
    public SymbolicRegressionSolution(ISymbolicRegressionModel model, IRegressionProblemData problemData)
      : base(model, problemData) {
      foreach (var node in model.SymbolicExpressionTree.Root.IterateNodesPrefix().OfType<SymbolicExpressionTreeTopLevelNode>())
        node.SetGrammar(null);

      Add(new Result(ModelLengthResultName, "Length of the symbolic regression model.", new IntValue()));
      Add(new Result(ModelDepthResultName, "Depth of the symbolic regression model.", new IntValue()));

      ResultCollection estimationLimitResults = new ResultCollection();
      estimationLimitResults.Add(new Result(EstimationLimitsResultName, "", new DoubleLimit()));
      estimationLimitResults.Add(new Result(TrainingUpperEstimationLimitHitsResultName, "", new IntValue()));
      estimationLimitResults.Add(new Result(TestUpperEstimationLimitHitsResultName, "", new IntValue()));
      estimationLimitResults.Add(new Result(TrainingLowerEstimationLimitHitsResultName, "", new IntValue()));
      estimationLimitResults.Add(new Result(TestLowerEstimationLimitHitsResultName, "", new IntValue()));
      estimationLimitResults.Add(new Result(TrainingNaNEvaluationsResultName, "", new IntValue()));
      estimationLimitResults.Add(new Result(TestNaNEvaluationsResultName, "", new IntValue()));
      Add(new Result(EstimationLimitsResultsResultName, "Results concerning the estimation limits of symbolic regression solution", estimationLimitResults));

      if (IntervalInterpreter.IsCompatible(Model.SymbolicExpressionTree))
        Add(new Result(ModelBoundsResultName, "Results concerning the derivation of symbolic regression solution", new IntervalCollection()));

      RecalculateResults();
    }

    public override IDeepCloneable Clone(Cloner cloner) {
      return new SymbolicRegressionSolution(this, cloner);
    }

    [StorableHook(HookType.AfterDeserialization)]
    private void AfterDeserialization() {
      if (!ContainsKey(EstimationLimitsResultsResultName)) {
        ResultCollection estimationLimitResults = new ResultCollection();
        estimationLimitResults.Add(new Result(EstimationLimitsResultName, "", new DoubleLimit()));
        estimationLimitResults.Add(new Result(TrainingUpperEstimationLimitHitsResultName, "", new IntValue()));
        estimationLimitResults.Add(new Result(TestUpperEstimationLimitHitsResultName, "", new IntValue()));
        estimationLimitResults.Add(new Result(TrainingLowerEstimationLimitHitsResultName, "", new IntValue()));
        estimationLimitResults.Add(new Result(TestLowerEstimationLimitHitsResultName, "", new IntValue()));
        estimationLimitResults.Add(new Result(TrainingNaNEvaluationsResultName, "", new IntValue()));
        estimationLimitResults.Add(new Result(TestNaNEvaluationsResultName, "", new IntValue()));
        Add(new Result(EstimationLimitsResultsResultName, "Results concerning the estimation limits of symbolic regression solution", estimationLimitResults));
        CalculateResults();
      }

      if (!ContainsKey(ModelBoundsResultName)) {
        if (IntervalInterpreter.IsCompatible(Model.SymbolicExpressionTree)) {
          Add(new Result(ModelBoundsResultName, "Results concerning the derivation of symbolic regression solution", new IntervalCollection()));
          CalculateResults();
        }
      }
    }

    protected override void RecalculateResults() {
      base.RecalculateResults();
      CalculateResults();
    }

    private void CalculateResults() {
      ModelLength = Model.SymbolicExpressionTree.Length;
      ModelDepth = Model.SymbolicExpressionTree.Depth;

      EstimationLimits.Lower = Model.LowerEstimationLimit;
      EstimationLimits.Upper = Model.UpperEstimationLimit;

      TrainingUpperEstimationLimitHits = EstimatedTrainingValues.Count(x => x.IsAlmost(Model.UpperEstimationLimit));
      TestUpperEstimationLimitHits = EstimatedTestValues.Count(x => x.IsAlmost(Model.UpperEstimationLimit));
      TrainingLowerEstimationLimitHits = EstimatedTrainingValues.Count(x => x.IsAlmost(Model.LowerEstimationLimit));
      TestLowerEstimationLimitHits = EstimatedTestValues.Count(x => x.IsAlmost(Model.LowerEstimationLimit));
      TrainingNaNEvaluations = Model.Interpreter.GetSymbolicExpressionTreeValues(Model.SymbolicExpressionTree, ProblemData.Dataset, ProblemData.TrainingIndices).Count(double.IsNaN);
      TestNaNEvaluations = Model.Interpreter.GetSymbolicExpressionTreeValues(Model.SymbolicExpressionTree, ProblemData.Dataset, ProblemData.TestIndices).Count(double.IsNaN);

      //Check if the tree contains unknown symbols for the interval calculation
      if (IntervalInterpreter.IsCompatible(Model.SymbolicExpressionTree))
        ModelBoundsCollection = CalculateModelIntervals(this);
    }

    private static IntervalCollection CalculateModelIntervals(ISymbolicRegressionSolution solution) {
      var intervalEvaluation = new IntervalCollection();
      var interpreter = new IntervalInterpreter();
      var problemData = solution.ProblemData;
      var model = solution.Model;
      var variableRanges = problemData.VariableRanges.GetReadonlyDictionary();

      intervalEvaluation.AddInterval($"Target {problemData.TargetVariable}", new Interval(variableRanges[problemData.TargetVariable].LowerBound, variableRanges[problemData.TargetVariable].UpperBound));
      intervalEvaluation.AddInterval("Model", interpreter.GetSymbolicExpressionTreeInterval(model.SymbolicExpressionTree, variableRanges));

      if (DerivativeCalculator.IsCompatible(model.SymbolicExpressionTree)) {
        foreach (var inputVariable in model.VariablesUsedForPrediction.OrderBy(v => v, new NaturalStringComparer())) {
          var derivedModel = DerivativeCalculator.Derive(model.SymbolicExpressionTree, inputVariable);
          var derivedResultInterval = interpreter.GetSymbolicExpressionTreeInterval(derivedModel, variableRanges);

          intervalEvaluation.AddInterval(" ∂f/∂" + inputVariable, new Interval(derivedResultInterval.LowerBound, derivedResultInterval.UpperBound));
        }
      }

      return intervalEvaluation;
    }

    public IEnumerable<double> EstimatedVariances => GetEstimatedVariances(ProblemData.AllIndices);

    public IEnumerable<double> EstimatedTrainingVariances => GetEstimatedVariances(ProblemData.TestIndices);

    public IEnumerable<double> EstimatedTestVariances => GetEstimatedVariances(ProblemData.TestIndices);


    public IEnumerable<double> GetEstimatedVariances(IEnumerable<int> rows) {
      return Model.GetEstimatedVariances(ProblemData.Dataset, rows);
    }
  }
}