[10214] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
| 3 | * Copyright (C) 2002-2013 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 |
|
---|
| 22 | using System.Linq;
|
---|
| 23 | using HeuristicLab.Core;
|
---|
| 24 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 25 | using HeuristicLab.Problems.DataAnalysis.Symbolic.Regression;
|
---|
| 26 |
|
---|
| 27 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
|
---|
| 28 | [Item("SymbolicDataAnalysisSolutionPruningOptimizer", "An operator which automatically removes nodes that have a negative impact from the tree model, optimizing the remaining constants.")]
|
---|
| 29 | [StorableClass]
|
---|
| 30 | public class SymbolicDataAnalysisRegressionSolutionPruningOptimizer : SymbolicDataAnalysisSolutionPruningOptimizer {
|
---|
| 31 | public override ISymbolicDataAnalysisSolution PruneAndOptimizeSolution(ISymbolicDataAnalysisSolution solution) {
|
---|
| 32 | var regressionSolution = (ISymbolicRegressionSolution)solution;
|
---|
| 33 | return PruneAndOptimizeRegressionSolution(regressionSolution);
|
---|
| 34 | }
|
---|
| 35 | /// <summary>
|
---|
| 36 | /// This method will walk all the levels of the symbolic regression solution model root starting from the deepest level and:
|
---|
| 37 | /// - it calculates the impact value of every originalNode on that level
|
---|
| 38 | /// - it prunes (replaces with a constant) the originalNode with the lowest negative impact value (0 or positive impacts are left unchanged)
|
---|
| 39 | /// - when no more nodes can be pruned, it moves on the upper level in the tree
|
---|
| 40 | /// - if the pruned and optimized solution is worse than the original solution (it can happen sometimes), then the original solution is returned
|
---|
| 41 | /// </summary>
|
---|
| 42 | /// <param name="solution"></param>
|
---|
| 43 | /// <returns></returns>
|
---|
| 44 | private ISymbolicRegressionSolution PruneAndOptimizeRegressionSolution(ISymbolicRegressionSolution solution) {
|
---|
| 45 | var calculator = new SymbolicRegressionSolutionImpactValuesCalculator();
|
---|
| 46 | var model = (ISymbolicRegressionModel)solution.Model;
|
---|
| 47 | var problemData = solution.ProblemData;
|
---|
| 48 | // get tree levels and iterate each level from the bottom up
|
---|
| 49 | var root = model.SymbolicExpressionTree.Root.GetSubtree(0).GetSubtree(0);
|
---|
| 50 | var levels = root.IterateNodesBreadth().GroupBy(root.GetBranchLevel).OrderByDescending(g => g.Key);
|
---|
| 51 |
|
---|
| 52 | OptimizeConstants(solution); // even if there are no negative impacts we still optimize the solution
|
---|
| 53 |
|
---|
| 54 | foreach (var level in levels) {
|
---|
| 55 | var nodes = level.ToArray();
|
---|
| 56 | double minImpact;
|
---|
| 57 | do {
|
---|
| 58 | minImpact = 0.0;
|
---|
| 59 | int minImpactIndex = -1;
|
---|
| 60 | for (int i = 0; i < nodes.Length; ++i) {
|
---|
| 61 | if (nodes[i] is ConstantTreeNode) continue;
|
---|
| 62 | var impact = calculator.CalculateImpactValue(model, nodes[i], problemData, problemData.TrainingIndices);
|
---|
| 63 | if (impact < minImpact) {
|
---|
| 64 | minImpact = impact;
|
---|
| 65 | minImpactIndex = i;
|
---|
| 66 | }
|
---|
| 67 | }
|
---|
| 68 | if (minImpact >= 0) continue;
|
---|
| 69 | var node = nodes[minImpactIndex];
|
---|
| 70 | var replacementValue = calculator.CalculateReplacementValue(model, node, problemData, problemData.TrainingIndices);
|
---|
| 71 | var constantNode = MakeConstantTreeNode(replacementValue);
|
---|
| 72 | ReplaceWithConstantNode(node, constantNode);
|
---|
| 73 | nodes[minImpactIndex] = constantNode;
|
---|
| 74 | OptimizeConstants(solution);
|
---|
| 75 | } while (minImpact < 0);
|
---|
| 76 | }
|
---|
| 77 |
|
---|
| 78 | var newSolution = (ISymbolicRegressionSolution)model.CreateRegressionSolution(problemData);
|
---|
| 79 | return newSolution.TrainingRSquared > solution.TrainingRSquared ? newSolution : solution;
|
---|
| 80 | }
|
---|
| 81 |
|
---|
| 82 | private static void OptimizeConstants(ISymbolicRegressionSolution solution) {
|
---|
| 83 | var model = solution.Model;
|
---|
| 84 | var problemData = solution.ProblemData;
|
---|
| 85 | SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(model.Interpreter, model.SymbolicExpressionTree, problemData, problemData.TrainingIndices, true, 50);
|
---|
| 86 | }
|
---|
| 87 | }
|
---|
| 88 | }
|
---|