Changeset 12745 for stable/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression

Timestamp:

07/12/15 11:23:06 (9 years ago)

Author:

gkronber

Message:

#2359, #2398: merged r12189,r12358,r12359,r12361,r12461,r12674,r12720,r12744 from trunk to stable

Location:

stable

Files:

• . (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis.Symbolic.Regression (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionPruningAnalyzer.cs (modified) (3 diffs)
• HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionPruningOperator.cs (modified) (3 diffs)
• HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionSolutionImpactValuesCalculator.cs (modified) (3 diffs)

stable/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionPruningAnalyzer.cs

@@ -24 +24 @@
 using HeuristicLab.Common;
 using HeuristicLab.Core;
+using HeuristicLab.Data;
 using HeuristicLab.Parameters;
 using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
@@ -31 +32 @@
   [StorableClass]
   public sealed class SymbolicRegressionPruningAnalyzer : SymbolicDataAnalysisSingleObjectivePruningAnalyzer {
-    private const string ImpactValuesCalculatorParameterName = "ImpactValuesCalculator";
     private const string PruningOperatorParameterName = "PruningOperator";
-    private SymbolicRegressionPruningAnalyzer(SymbolicRegressionPruningAnalyzer original, Cloner cloner)
-      : base(original, cloner) {
+    public IValueParameter<SymbolicRegressionPruningOperator> PruningOperatorParameter {
+      get { return (IValueParameter<SymbolicRegressionPruningOperator>)Parameters[PruningOperatorParameterName]; }
     }
-    public override IDeepCloneable Clone(Cloner cloner) {
-      return new SymbolicRegressionPruningAnalyzer(this, cloner);
+
+    protected override SymbolicDataAnalysisExpressionPruningOperator PruningOperator {
+      get { return PruningOperatorParameter.Value; }
     }
+
+    private SymbolicRegressionPruningAnalyzer(SymbolicRegressionPruningAnalyzer original, Cloner cloner) : base(original, cloner) { }
+    public override IDeepCloneable Clone(Cloner cloner) { return new SymbolicRegressionPruningAnalyzer(this, cloner); }
 
     [StorableConstructor]
@@ -44 +48 @@
 
     public SymbolicRegressionPruningAnalyzer() {
-      Parameters.Add(new ValueParameter<SymbolicDataAnalysisSolutionImpactValuesCalculator>(ImpactValuesCalculatorParameterName, "The impact values calculator", new SymbolicRegressionSolutionImpactValuesCalculator()));
-      Parameters.Add(new ValueParameter<SymbolicDataAnalysisExpressionPruningOperator>(PruningOperatorParameterName, "The operator used to prune trees", new SymbolicRegressionPruningOperator()));
+      Parameters.Add(new ValueParameter<SymbolicRegressionPruningOperator>(PruningOperatorParameterName, "The operator used to prune trees", new SymbolicRegressionPruningOperator(new SymbolicRegressionSolutionImpactValuesCalculator())));
+    }
+
+    [StorableHook(HookType.AfterDeserialization)]
+    private void AfterDeserialization() {
+      // BackwardsCompatibility3.3
+
+      #region Backwards compatible code, remove with 3.4
+      if (Parameters.ContainsKey(PruningOperatorParameterName)) {
+        var oldParam = Parameters[PruningOperatorParameterName] as ValueParameter<SymbolicDataAnalysisExpressionPruningOperator>;
+        if (oldParam != null) {
+          Parameters.Remove(oldParam);
+          Parameters.Add(new ValueParameter<SymbolicRegressionPruningOperator>(PruningOperatorParameterName, "The operator used to prune trees", new SymbolicRegressionPruningOperator(new SymbolicRegressionSolutionImpactValuesCalculator())));
+        }
+      } else {
+        // not yet contained
+        Parameters.Add(new ValueParameter<SymbolicRegressionPruningOperator>(PruningOperatorParameterName, "The operator used to prune trees", new SymbolicRegressionPruningOperator(new SymbolicRegressionSolutionImpactValuesCalculator())));
+      }
+
+
+      if (Parameters.ContainsKey("PruneOnlyZeroImpactNodes")) {
+        PruningOperator.PruneOnlyZeroImpactNodes = ((IFixedValueParameter<BoolValue>)Parameters["PruneOnlyZeroImpactNodes"]).Value.Value;
+        Parameters.Remove(Parameters["PruneOnlyZeroImpactNodes"]);
+      }
+      if (Parameters.ContainsKey("ImpactThreshold")) {
+        PruningOperator.NodeImpactThreshold = ((IFixedValueParameter<DoubleValue>)Parameters["ImpactThreshold"]).Value.Value;
+        Parameters.Remove(Parameters["ImpactThreshold"]);
+      }
+      if (Parameters.ContainsKey("ImpactValuesCalculator")) {
+        PruningOperator.ImpactValuesCalculator = ((ValueParameter<SymbolicDataAnalysisSolutionImpactValuesCalculator>)Parameters["ImpactValuesCalculator"]).Value;
+        Parameters.Remove(Parameters["ImpactValuesCalculator"]);
+      }
+
+      #endregion
     }
   }

stable/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionPruningOperator.cs

@@ -22 +22 @@
 #endregion
 
+using System.Collections.Generic;
 using System.Linq;
 using HeuristicLab.Common;
 using HeuristicLab.Core;
+using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
 using HeuristicLab.Parameters;
 using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
@@ -32 +34 @@
   [Item("SymbolicRegressionPruningOperator", "An operator which prunes symbolic regression trees.")]
   public class SymbolicRegressionPruningOperator : SymbolicDataAnalysisExpressionPruningOperator {
-    private const string ImpactValuesCalculatorParameterName = "ImpactValuesCalculator";
+    private const string EvaluatorParameterName = "Evaluator";
+
+    #region parameter properties
+    public ILookupParameter<ISymbolicRegressionSingleObjectiveEvaluator> EvaluatorParameter {
+      get { return (ILookupParameter<ISymbolicRegressionSingleObjectiveEvaluator>)Parameters[EvaluatorParameterName]; }
+    }
+    #endregion
 
     protected SymbolicRegressionPruningOperator(SymbolicRegressionPruningOperator original, Cloner cloner)
@@ -44 +52 @@
     protected SymbolicRegressionPruningOperator(bool deserializing) : base(deserializing) { }
 
-    public SymbolicRegressionPruningOperator() {
-      var impactValuesCalculator = new SymbolicRegressionSolutionImpactValuesCalculator();
-      Parameters.Add(new ValueParameter<ISymbolicDataAnalysisSolutionImpactValuesCalculator>(ImpactValuesCalculatorParameterName, "The impact values calculator to be used for figuring out the node impacts.", impactValuesCalculator));
+    public SymbolicRegressionPruningOperator(ISymbolicDataAnalysisSolutionImpactValuesCalculator impactValuesCalculator)
+      : base(impactValuesCalculator) {
+      Parameters.Add(new LookupParameter<ISymbolicRegressionSingleObjectiveEvaluator>(EvaluatorParameterName));
     }
 
-    protected override ISymbolicDataAnalysisModel CreateModel() {
-      return new SymbolicRegressionModel(SymbolicExpressionTree, Interpreter, EstimationLimits.Lower, EstimationLimits.Upper);
+    [StorableHook(HookType.AfterDeserialization)]
+    private void AfterDeserialization() {
+      // BackwardsCompatibility3.3
+      #region Backwards compatible code, remove with 3.4
+      base.ImpactValuesCalculator = new SymbolicRegressionSolutionImpactValuesCalculator();
+      if (!Parameters.ContainsKey(EvaluatorParameterName)) {
+        Parameters.Add(new LookupParameter<ISymbolicRegressionSingleObjectiveEvaluator>(EvaluatorParameterName));
+      }
+      #endregion
+    }
+
+    protected override ISymbolicDataAnalysisModel CreateModel(ISymbolicExpressionTree tree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, IDataAnalysisProblemData problemData, DoubleLimit estimationLimits) {
+      return new SymbolicRegressionModel(tree, interpreter, estimationLimits.Lower, estimationLimits.Upper);
     }
 
     protected override double Evaluate(IDataAnalysisModel model) {
-      var regressionModel = (IRegressionModel)model;
-      var regressionProblemData = (IRegressionProblemData)ProblemData;
-      var trainingIndices = Enumerable.Range(FitnessCalculationPartition.Start, FitnessCalculationPartition.Size);
-      var estimatedValues = regressionModel.GetEstimatedValues(ProblemData.Dataset, trainingIndices); // also bounds the values
-      var targetValues = ProblemData.Dataset.GetDoubleValues(regressionProblemData.TargetVariable, trainingIndices);
-      OnlineCalculatorError errorState;
-      var quality = OnlinePearsonsRCalculator.Calculate(targetValues, estimatedValues, out errorState);
-      if (errorState != OnlineCalculatorError.None) return double.NaN;
-      return quality*quality;
+      var regressionModel = (ISymbolicRegressionModel)model;
+      var regressionProblemData = (IRegressionProblemData)ProblemDataParameter.ActualValue;
+      var evaluator = EvaluatorParameter.ActualValue;
+      var fitnessEvaluationPartition = FitnessCalculationPartitionParameter.ActualValue;
+      var rows = Enumerable.Range(fitnessEvaluationPartition.Start, fitnessEvaluationPartition.Size);
+      return evaluator.Evaluate(this.ExecutionContext, regressionModel.SymbolicExpressionTree, regressionProblemData, rows);
+    }
+
+    public static ISymbolicExpressionTree Prune(ISymbolicExpressionTree tree, SymbolicRegressionSolutionImpactValuesCalculator impactValuesCalculator, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, IRegressionProblemData problemData, DoubleLimit estimationLimits, IEnumerable<int> rows, double nodeImpactThreshold = 0.0, bool pruneOnlyZeroImpactNodes = false) {
+      var clonedTree = (ISymbolicExpressionTree)tree.Clone();
+      var model = new SymbolicRegressionModel(clonedTree, interpreter, estimationLimits.Lower, estimationLimits.Upper);
+      var nodes = clonedTree.Root.GetSubtree(0).GetSubtree(0).IterateNodesPrefix().ToList(); // skip the nodes corresponding to the ProgramRootSymbol and the StartSymbol
+
+      double qualityForImpactsCalculation = double.NaN; // pass a NaN value initially so the impact calculator will calculate the quality
+
+      for (int i = 0; i < nodes.Count; ++i) {
+        var node = nodes[i];
+        if (node is ConstantTreeNode) continue;
+
+        double impactValue, replacementValue;
+        double newQualityForImpactsCalculation;
+        impactValuesCalculator.CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, out newQualityForImpactsCalculation, qualityForImpactsCalculation);
+
+        if (pruneOnlyZeroImpactNodes && !impactValue.IsAlmost(0.0)) continue;
+        if (!pruneOnlyZeroImpactNodes && impactValue > nodeImpactThreshold) continue;
+
+        var constantNode = (ConstantTreeNode)node.Grammar.GetSymbol("Constant").CreateTreeNode();
+        constantNode.Value = replacementValue;
+
+        ReplaceWithConstant(node, constantNode);
+        i += node.GetLength() - 1; // skip subtrees under the node that was folded
+
+        qualityForImpactsCalculation = newQualityForImpactsCalculation;
+      }
+      return model.SymbolicExpressionTree;
     }
   }

stable/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SymbolicRegressionSolutionImpactValuesCalculator.cs

@@ -48 +48 @@
     }
 
-    public override double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double originalQuality = double.NaN) {
-      double impactValue, replacementValue;
-      CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, originalQuality);
+    public override double CalculateImpactValue(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node, IDataAnalysisProblemData problemData, IEnumerable<int> rows, double qualityForImpactsCalculation = double.NaN) {
+      double impactValue, replacementValue, newQualityForImpactsCalculation;
+      CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, out newQualityForImpactsCalculation, qualityForImpactsCalculation);
       return impactValue;
     }
 
     public override void CalculateImpactAndReplacementValues(ISymbolicDataAnalysisModel model, ISymbolicExpressionTreeNode node,
-      IDataAnalysisProblemData problemData, IEnumerable<int> rows, out double impactValue, out double replacementValue,
-      double originalQuality = Double.NaN) {
+      IDataAnalysisProblemData problemData, IEnumerable<int> rows, out double impactValue, out double replacementValue, out double newQualityForImpactsCalculation,
+      double qualityForImpactsCalculation = Double.NaN) {
       var regressionModel = (ISymbolicRegressionModel)model;
       var regressionProblemData = (IRegressionProblemData)problemData;
@@ -64 +64 @@
 
       OnlineCalculatorError errorState;
-      if (double.IsNaN(originalQuality)) {
-        var originalValues = regressionModel.GetEstimatedValues(dataset, rows);
-        originalQuality = OnlinePearsonsRCalculator.Calculate(targetValues, originalValues, out errorState);
-        if (errorState != OnlineCalculatorError.None) originalQuality = 0.0;
-      }
+      if (double.IsNaN(qualityForImpactsCalculation))
+        qualityForImpactsCalculation = CalculateQualityForImpacts(regressionModel, regressionProblemData, rows);
 
       replacementValue = CalculateReplacementValue(regressionModel, node, regressionProblemData, rows);
@@ -83 +80 @@
 
       var estimatedValues = tempModel.GetEstimatedValues(dataset, rows);
-      double newQuality = OnlinePearsonsRCalculator.Calculate(targetValues, estimatedValues, out errorState);
-      if (errorState != OnlineCalculatorError.None) newQuality = 0.0;
+      double r = OnlinePearsonsRCalculator.Calculate(targetValues, estimatedValues, out errorState);
+      if (errorState != OnlineCalculatorError.None) r = 0.0;
+      newQualityForImpactsCalculation = r * r;
 
-      impactValue = (originalQuality*originalQuality) - (newQuality*newQuality);
+      impactValue = qualityForImpactsCalculation - newQualityForImpactsCalculation;
+    }
+
+    public static double CalculateQualityForImpacts(ISymbolicRegressionModel model, IRegressionProblemData problemData, IEnumerable<int> rows) {
+      var estimatedValues = model.GetEstimatedValues(problemData.Dataset, rows); // also bounds the values
+      var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
+      OnlineCalculatorError errorState;
+      var r = OnlinePearsonsRCalculator.Calculate(targetValues, estimatedValues, out errorState);
+      var quality = r * r;
+      if (errorState != OnlineCalculatorError.None) return double.NaN;
+      return quality;
     }
   }