Changeset 12720 for trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic

Timestamp:

07/10/15 15:41:09 (9 years ago)

Author:

bburlacu

Message:

#2359: Changed the impact calculators so that the quality value necessary for impacts calculation is calculated with a separate method. Refactored the CalculateImpactAndReplacementValues method to return the new quality in an out-parameter (adjusted method signature in interface accordingly). Added Evaluate method to the regression and classification pruning operators that re-evaluates the tree using the problem evaluator after pruning was performed.

Location:

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4

Files:

• Interfaces/ISymbolicDataAnalysisImpactValuesCalculator.cs (modified) (1 diff)
• SymbolicDataAnalysisExpressionPruningOperator.cs (modified) (7 diffs)
• SymbolicDataAnalysisSolutionImpactValuesCalculator.cs (modified) (1 diff)

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Interfaces/ISymbolicDataAnalysisImpactValuesCalculator.cs

@@ -6 +6 @@
   public interface ISymbolicDataAnalysisSolutionImpactValuesCalculator : IItem {
     double CalculateReplacementValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows);
-    double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double originalQuality = double.NaN);
+    double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double qualityForImpactsCalculation = double.NaN);
     void CalculateImpactAndReplacementValues(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData,
-      IEnumerable<int> rows, out double impactValue, out double replacementValue, double originalQuality = double.NaN);
+      IEnumerable<int> rows, out double impactValue, out double replacementValue, out double newQualityForImpactsCalculation, double qualityForImpactsCalculation = double.NaN);
   }
 }

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/SymbolicDataAnalysisExpressionPruningOperator.cs

@@ -49 +49 @@
     private const string EstimationLimitsParameterName = "EstimationLimits";
     private const string InterpreterParameterName = "SymbolicExpressionTreeInterpreter";
+    private const string ApplyLinearScalingParameterName = "ApplyLinearScaling";
     #endregion
 
@@ -88 +89 @@
       get { return (ILookupParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>)Parameters[InterpreterParameterName]; }
     }
+    public ILookupParameter<BoolValue> ApplyLinearScalingParameter {
+      get { return (ILookupParameter<BoolValue>)Parameters[ApplyLinearScalingParameterName]; }
+    }
     #endregion
 
@@ -124 +128 @@
       Parameters.Add(new LookupParameter<ISymbolicExpressionTree>(SymbolicExpressionTreeParameterName));
       Parameters.Add(new LookupParameter<DoubleValue>(QualityParameterName));
+      Parameters.Add(new LookupParameter<BoolValue>(ApplyLinearScalingParameterName));
       Parameters.Add(new ValueParameter<ISymbolicDataAnalysisSolutionImpactValuesCalculator>(ImpactValuesCalculatorParameterName, impactValuesCalculator));
       #endregion
@@ -141 +146 @@
       var model = CreateModel(tree, interpreter, problemData, estimationLimits);
       var nodes = tree.Root.GetSubtree(0).GetSubtree(0).IterateNodesPrefix().ToList();
-      var rows = Enumerable.Range(fitnessCalculationPartition.Start, fitnessCalculationPartition.Size);
+      var rows = Enumerable.Range(fitnessCalculationPartition.Start, fitnessCalculationPartition.Size).ToList();
       var prunedSubtrees = 0;
       var prunedTrees = 0;
       var prunedNodes = 0;
 
-      double quality = Evaluate(model);
+      double qualityForImpactsCalculation = double.NaN;
 
       for (int i = 0; i < nodes.Count; ++i) {
@@ -153 +158 @@
 
         double impactValue, replacementValue;
-        ImpactValuesCalculator.CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, quality);
+        double newQualityForImpacts;
+        ImpactValuesCalculator.CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, out newQualityForImpacts, qualityForImpactsCalculation);
 
         if (PruneOnlyZeroImpactNodes && !impactValue.IsAlmost(0.0)) continue;
@@ -165 +171 @@
         i += length - 1; // skip subtrees under the node that was folded
 
-        quality -= impactValue;
         prunedSubtrees++;
         prunedNodes += length;
+
+        qualityForImpactsCalculation = newQualityForImpacts;
       }
 
@@ -174 +181 @@
       PrunedTreesParameter.ActualValue = new IntValue(prunedTrees);
       PrunedNodesParameter.ActualValue = new IntValue(prunedNodes);
-      QualityParameter.ActualValue.Value = quality;
+
+      if (prunedSubtrees > 0) // if nothing was pruned then there's no need to re-evaluate the tree
+        QualityParameter.ActualValue.Value = Evaluate(model);
 
       return base.Apply();

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/SymbolicDataAnalysisSolutionImpactValuesCalculator.cs

@@ -37 +37 @@
     protected SymbolicDataAnalysisSolutionImpactValuesCalculator(bool deserializing) : base(deserializing) { }
     public abstract double CalculateReplacementValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows);
-    public abstract double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double originalQuality = double.NaN);
-    public abstract void CalculateImpactAndReplacementValues(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, out double impactValue, out double replacementValue, double originalQuality = double.NaN);
+    public abstract double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double qualityForImpactsCalculation = double.NaN);
+    public abstract void CalculateImpactAndReplacementValues(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, out double impactValue, out double replacementValue, out double newQualityForImpactsCalculation, double qualityForImpactsCalculation = double.NaN);
 
     protected static double CalculateReplacementValue(ISymbolicExpressionTreeNode node, ISymbolicExpressionTree sourceTree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter,