Changeset 1796 for trunk/sources/HeuristicLab.GP.StructureIdentification

Timestamp:

05/14/09 13:46:57 (16 years ago)

Author:

gkronber

Message:

Refactored GP evaluation to make it possible to use different evaluators to interpret function trees. #615 (Evaluation of HL3 function trees should be equivalent to evaluation in HL2)

Location:

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3

Files:

• BakedTreeEvaluator.cs (modified) (4 diffs)
• Evaluators/CoefficientOfDeterminationEvaluator.cs (modified) (2 diffs)
• Evaluators/EarlyStoppingMeanSquaredErrorEvaluator.cs (modified) (2 diffs)
• Evaluators/GPEvaluatorBase.cs (modified) (5 diffs)
• Evaluators/MeanAbsolutePercentageErrorEvaluator.cs (modified) (1 diff)
• Evaluators/MeanAbsolutePercentageOfRangeErrorEvaluator.cs (modified) (1 diff)
• Evaluators/MeanSquaredErrorEvaluator.cs (modified) (1 diff)
• Evaluators/SimpleEvaluator.cs (modified) (2 diffs)
• Evaluators/UncertainMeanSquaredErrorEvaluator.cs (modified) (2 diffs)
• Evaluators/VarianceAccountedForEvaluator.cs (modified) (2 diffs)
• HeuristicLab.GP.StructureIdentification-3.3.csproj (modified) (1 diff)
• ITreeEvaluator.cs (added)
• TreeEvaluatorInjector.cs (added)

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/BakedTreeEvaluator.cs ¶

@@ -34 +34 @@
   /// Not thread-safe!
   /// </summary>
-  public class BakedTreeEvaluator {
+  public class BakedTreeEvaluator : ItemBase, ITreeEvaluator {
     private const double EPSILON = 1.0e-7;
     private double estimatedValueMax;
@@ -53 +53 @@
     private int sampleIndex;
 
-    public void ResetEvaluator(BakedFunctionTree functionTree, Dataset dataset, int targetVariable, int start, int end, double punishmentFactor) {
+    public void ResetEvaluator(Dataset dataset, int targetVariable, int start, int end, double punishmentFactor) {
       this.dataset = dataset;
       double maximumPunishment = punishmentFactor * dataset.GetRange(targetVariable, start, end);
@@ -62 +62 @@
       estimatedValueMax = targetMean + maximumPunishment;
 
-      List<LightWeightFunction> linearRepresentation = functionTree.LinearRepresentation;
-      codeArr = new Instr[linearRepresentation.Count];
-      int i = 0;
-      foreach (LightWeightFunction f in linearRepresentation) {
-        codeArr[i++] = TranslateToInstr(f);
-      }
     }
 
@@ -94 +88 @@
     }
 
-    public double Evaluate(int sampleIndex) {
+    public double Evaluate(IFunctionTree functionTree, int sampleIndex) {
+      BakedFunctionTree bakedTree = functionTree as BakedFunctionTree;
+      if (bakedTree == null) throw new ArgumentException("BakedTreeEvaluator can only evaluate BakedFunctionTrees");
+
+      List<LightWeightFunction> linearRepresentation = bakedTree.LinearRepresentation;
+      codeArr = new Instr[linearRepresentation.Count];
+      int i = 0;
+      foreach (LightWeightFunction f in linearRepresentation) {
+        codeArr[i++] = TranslateToInstr(f);
+      }
+
       PC = 0;
       this.sampleIndex = sampleIndex;

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/CoefficientOfDeterminationEvaluator.cs ¶

@@ -42 +42 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double errorsSquaredSum = 0.0;
       double originalDeviationTotalSumOfSquares = 0.0;
       double targetMean = dataset.GetMean(targetVariable, start, end);
+
+      double originalSum = 0.0;
+      int n = 0;
       for (int sample = start; sample < end; sample++) {
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
         if (updateTargetValues) {
@@ -56 +59 @@
           errorsSquaredSum += error * error;
 
-          double origDeviation = original - targetMean;
-          originalDeviationTotalSumOfSquares += origDeviation * origDeviation;
+          originalSum += original;
+          n++;
+        }
+      }
+
+      double originalMean = originalSum / n;
+      for(int sample = start; sample < end; sample++){
+        double original = dataset.GetValue(sample, targetVariable);
+        if (!double.IsNaN(original) && !double.IsInfinity(original)) {
+          original = original - originalMean;
+          original = original * original;
+          originalDeviationTotalSumOfSquares += original;
         }
       }

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/EarlyStoppingMeanSquaredErrorEvaluator.cs ¶

@@ -44 +44 @@
 
     // evaluates the function-tree for the given target-variable and the whole dataset and returns the MSE
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double qualityLimit = GetVariableValue<DoubleData>("QualityLimit", scope, false).Data;
       DoubleData mse = GetVariableValue<DoubleData>("MSE", scope, false, false);
@@ -56 +56 @@
       int n = 0;
       for (int sample = start; sample < end; sample++) {
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
         if (updateTargetValues) {

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/GPEvaluatorBase.cs ¶

@@ -33 +33 @@
     public GPEvaluatorBase()
       : base() {
+      AddVariableInfo(new VariableInfo("TreeEvaluator", "The evaluator that should be used to evaluate the expression tree", typeof(ITreeEvaluator), VariableKind.In));
       AddVariableInfo(new VariableInfo("FunctionTree", "The function tree that should be evaluated", typeof(IFunctionTree), VariableKind.In));
       AddVariableInfo(new VariableInfo("TreeSize", "Size (number of nodes) of the tree to evaluate", typeof(IntData), VariableKind.In));
@@ -48 +49 @@
       int targetVariable = GetVariableValue<IntData>("TargetVariable", scope, true).Data;
       Dataset dataset = GetVariableValue<Dataset>("Dataset", scope, true);
-      BakedFunctionTree functionTree = GetVariableValue<BakedFunctionTree>("FunctionTree", scope, true);
+      IFunctionTree functionTree = GetVariableValue<IFunctionTree>("FunctionTree", scope, true);
       double punishmentFactor = GetVariableValue<DoubleData>("PunishmentFactor", scope, true).Data;
       int treeSize = scope.GetVariableValue<IntData>("TreeSize", false).Data;
@@ -55 +56 @@
       int end = GetVariableValue<IntData>("SamplesEnd", scope, true).Data;
       bool useEstimatedValues = GetVariableValue<BoolData>("UseEstimatedTargetValue", scope, true).Data;
+      ITreeEvaluator evaluator = GetVariableValue<ITreeEvaluator>("TreeEvaluator", scope, true);
+
       double[] backupValues = null;
       // prepare for autoregressive modelling by saving the original values of the target-variable to a backup array
@@ -65 +68 @@
       }
 
-      // initialize and reset the evaluator
-      BakedTreeEvaluator evaluator = new BakedTreeEvaluator();
-      evaluator.ResetEvaluator(functionTree, dataset, targetVariable, start, end, punishmentFactor);
-
-      Evaluate(scope, evaluator, dataset, targetVariable, start, end, useEstimatedValues);
+      Evaluate(scope, evaluator, functionTree, dataset, targetVariable, start, end, useEstimatedValues);
 
       // restore the values of the target variable from the backup array if necessary
@@ -83 +82 @@
     }
 
-    public abstract void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues);
+    public abstract void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues);
   }
 }

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/MeanAbsolutePercentageErrorEvaluator.cs ¶

@@ -43 +43 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double errorsSum = 0.0;
       int n = 0;
       for (int sample = start; sample < end; sample++) {
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
 

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/MeanAbsolutePercentageOfRangeErrorEvaluator.cs ¶

@@ -43 +43 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double errorsSum = 0.0;
       int n = 0;
       double range = dataset.GetRange(targetVariable, start, end);
       for (int sample = start; sample < end; sample++) {
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
 

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/MeanSquaredErrorEvaluator.cs ¶

@@ -43 +43 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double errorsSquaredSum = 0;
       int n = 0;
       for (int sample = start; sample < end; sample++) {
         double original = dataset.GetValue(sample, targetVariable);
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         if (updateTargetValues) {
           dataset.SetValue(sample, targetVariable, estimated);

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/SimpleEvaluator.cs ¶

@@ -36 +36 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       ItemList values = GetVariableValue<ItemList>("Values", scope, false, false);
       if (values == null) {
@@ -50 +50 @@
       for (int sample = start; sample < end; sample++) {
         ItemList row = new ItemList();
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
         if (updateTargetValues) {

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/UncertainMeanSquaredErrorEvaluator.cs ¶

@@ -50 +50 @@
 
     // evaluates the function-tree for the given target-variable and the whole dataset and returns the MSE
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, HeuristicLab.DataAnalysis.Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       double qualityLimit = GetVariableValue<DoubleData>("QualityLimit", scope, false).Data;
       int minSamples = GetVariableValue<IntData>("MinEvaluatedSamples", scope, true).Data;
@@ -78 +78 @@
       int n = 0;
       for (int sample = 0; sample < rows; sample++) {
-        double estimated = evaluator.Evaluate(indexes[sample]);
+        double estimated = evaluator.Evaluate(tree, indexes[sample]);
         double original = dataset.GetValue(indexes[sample], targetVariable);
         if (!double.IsNaN(original) && !double.IsInfinity(original)) {

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/Evaluators/VarianceAccountedForEvaluator.cs ¶

@@ -53 +53 @@
     }
 
-    public override void Evaluate(IScope scope, BakedTreeEvaluator evaluator, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
+    public override void Evaluate(IScope scope, ITreeEvaluator evaluator, IFunctionTree tree, Dataset dataset, int targetVariable, int start, int end, bool updateTargetValues) {
       int nSamples = end - start;
       double[] errors = new double[nSamples];
       double[] originalTargetVariableValues = new double[nSamples];
       for (int sample = start; sample < end; sample++) {
-        double estimated = evaluator.Evaluate(sample);
+        double estimated = evaluator.Evaluate(tree, sample);
         double original = dataset.GetValue(sample, targetVariable);
         if (updateTargetValues) {
@@ -66 +66 @@
           errors[sample - start] = original - estimated;
           originalTargetVariableValues[sample - start] = original;
+        } else {
+          errors[sample - start] = double.NaN;
+          originalTargetVariableValues[sample - start] = double.NaN;
         }
       }

trunk/sources/HeuristicLab.GP.StructureIdentification/3.3/HeuristicLab.GP.StructureIdentification-3.3.csproj ¶

@@ -89 +89 @@
     <Compile Include="Constant.cs" />
     <Compile Include="AlgorithmBase.cs" />
+    <Compile Include="TreeEvaluatorInjector.cs" />
+    <Compile Include="ITreeEvaluator.cs" />
     <Compile Include="Evaluators\MeanAbsolutePercentageOfRangeErrorEvaluator.cs" />
     <Compile Include="FunctionLibraryInjector.cs" />