source: branches/HeuristicLab.EvolutionaryTracking/HeuristicLab.EvolutionaryTracking/3.4/Analyzers/SymbolicDataAnalysisSolutionPruningOptimizer.cs @ 9963

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2013 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.Linq;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Operators;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
using HeuristicLab.Problems.DataAnalysis.Symbolic;
using HeuristicLab.Problems.DataAnalysis.Symbolic.Regression;

namespace HeuristicLab.EvolutionaryTracking {
  [Item("SymbolicDataAnalysisSolutionPruningOptimizer", "An operator which automatically removes nodes that have a negative impact from the tree model, optimizing the remaining constants.")]
  [StorableClass]
  public class SymbolicDataAnalysisSolutionPruningOptimizer : SingleSuccessorOperator, IAnalyzer {
    private const string ResultCollectionParameterName = "Results";
    private const string TrainingBestSolutionParameterName = "Best training solution";

    #region parameter properties
    public ILookupParameter<ResultCollection> ResultCollectionParameter {
      get { return (ILookupParameter<ResultCollection>)Parameters[ResultCollectionParameterName]; }
    }
    #endregion
    #region properties
    public ResultCollection ResultCollection {
      get { return ResultCollectionParameter.ActualValue; }
    }
    #endregion

    [StorableConstructor]
    protected SymbolicDataAnalysisSolutionPruningOptimizer(bool deserializing) : base(deserializing) { }

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

    protected SymbolicDataAnalysisSolutionPruningOptimizer(SymbolicDataAnalysisSolutionPruningOptimizer original, Cloner cloner)
      : base(original, cloner) {
    }

    public SymbolicDataAnalysisSolutionPruningOptimizer() {
      Parameters.Add(new LookupParameter<ResultCollection>(ResultCollectionParameterName));
    }

    public override IOperation Apply() {
      var solution = (ISymbolicRegressionSolution)ResultCollection[TrainingBestSolutionParameterName].Value;
      ResultCollection[TrainingBestSolutionParameterName].Value = PruneAndOptimizeSolution(solution);
      return base.Apply();
    }

    public ISymbolicDataAnalysisSolution PruneAndOptimizeSolution(ISymbolicRegressionSolution solution) {
      return PruneAndOptimizeRegressionSolution(solution);
    }

    private static ISymbolicRegressionSolution PruneAndOptimizeRegressionSolution(ISymbolicRegressionSolution solution) {
      var calculator = new SymbolicRegressionSolutionImpactValuesCalculator();
      var model = solution.Model;
      var problemData = solution.ProblemData;
      // get tree levels and iterate each level from the bottom up
      var root = model.SymbolicExpressionTree.Root.GetSubtree(0).GetSubtree(0);
      var levels = root.IterateNodesBreadth().GroupBy(root.GetBranchLevel).OrderByDescending(g => g.Key);

      foreach (var level in levels) {
        bool stop = false;
        while (!stop) {
          var tuple = level.Select(n => new { Node = n, Impact = calculator.CalculateImpactValue(model, n, problemData, problemData.TrainingIndices) }).OrderBy(x => x.Impact).First();
          var node = tuple.Node;
          var impact = tuple.Impact;
          if (impact < 0) {
            var replacementValue = calculator.CalculateReplacementValue(model, node, problemData,
              problemData.TrainingIndices);
            var constantNode = MakeConstantTreeNode(replacementValue);
            var parent = node.Parent;
            var index = parent.IndexOfSubtree(node);
            parent.RemoveSubtree(index);
            parent.InsertSubtree(index, constantNode);
            SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(model.Interpreter,
              model.SymbolicExpressionTree, problemData, problemData.TrainingIndices, true, 50,
              model.UpperEstimationLimit,
              model.LowerEstimationLimit);
            continue;
          }
          stop = true;
        }
      }
      var newSolution = (ISymbolicRegressionSolution)model.CreateRegressionSolution(problemData);
      return newSolution.TrainingRSquared > solution.TrainingRSquared ? newSolution : solution;
    }

    public bool EnabledByDefault {
      get { return true; }
    }

    private static ConstantTreeNode MakeConstantTreeNode(double value) {
      var constant = new Constant { MinValue = value - 1, MaxValue = value + 1 };
      var constantTreeNode = (ConstantTreeNode)constant.CreateTreeNode();
      constantTreeNode.Value = value;
      return constantTreeNode;
    }
  }
}