Changeset 15883 for branches/2886_SymRegGrammarEnumeration/HeuristicLab.Algorithms.DataAnalysis.SymRegGrammarEnumeration/Analysis/RSquaredEvaluator.cs

Timestamp:

04/04/18 16:23:55 (6 years ago)

Author:

lkammere

Message:

#2886: Priorize phrases whose (fully expanded) terms result in high R².

File:

• branches/2886_SymRegGrammarEnumeration/HeuristicLab.Algorithms.DataAnalysis.SymRegGrammarEnumeration/Analysis/RSquaredEvaluator.cs (modified) (4 diffs)

branches/2886_SymRegGrammarEnumeration/HeuristicLab.Algorithms.DataAnalysis.SymRegGrammarEnumeration/Analysis/RSquaredEvaluator.cs

@@ -13 +13 @@
 namespace HeuristicLab.Algorithms.DataAnalysis.SymRegGrammarEnumeration {
   public class RSquaredEvaluator : Item, IGrammarEnumerationAnalyzer {
-    private readonly string BestTrainingQualityResultName = "Best R² (Training)";
-    private readonly string BestTrainingModelResultName = "Best model (Training)";
-    private readonly string BestTrainingSolutionResultName = "Best solution (Training)";
+    public static readonly string BestTrainingQualityResultName = "Best R² (Training)";
+    public static readonly string BestTrainingModelResultName = "Best model (Training)";
+    public static readonly string BestTrainingSolutionResultName = "Best solution (Training)";
 
-    private readonly ISymbolicDataAnalysisExpressionTreeInterpreter expressionTreeLinearInterpreter = new SymbolicDataAnalysisExpressionTreeLinearInterpreter();
+    private static readonly ISymbolicDataAnalysisExpressionTreeInterpreter expressionTreeLinearInterpreter = new SymbolicDataAnalysisExpressionTreeLinearInterpreter();
 
     public bool OptimizeConstants { get; set; }
@@ -73 +73 @@
       Debug.Assert(SymbolicRegressionConstantOptimizationEvaluator.CanOptimizeConstants(tree));
 
-      // TODO: Initialize constant values randomly
-      // TODO: Restarts
+      double r2 = Evaluate(problemData, tree, OptimizeConstants);
 
-      double r2;
+      var bestR2Result = (DoubleValue)algorithm.Results[BestTrainingQualityResultName].Value;
+      bool better = r2 > bestR2Result.Value;
+      bool equallyGood = r2.IsAlmost(bestR2Result.Value);
+      bool shorter = false;
 
-      SymbolicRegressionModel model = new SymbolicRegressionModel(
+      if (!better && equallyGood) {
+        shorter = algorithm.BestTrainingSentence != null &&
+          algorithm.Grammar.GetComplexity(algorithm.BestTrainingSentence) > algorithm.Grammar.GetComplexity(symbolString);
+      }
+      if (better || (equallyGood && shorter)) {
+        bestR2Result.Value = r2;
+
+        SymbolicRegressionModel model = new SymbolicRegressionModel(
           problemData.TargetVariable,
           tree,
           expressionTreeLinearInterpreter);
 
-      if (OptimizeConstants) {
+        algorithm.Results.AddOrUpdateResult(BestTrainingModelResultName, model);
+
+        algorithm.BestTrainingSentence = symbolString;
+      }
+    }
+
+    public static double Evaluate(IRegressionProblemData problemData, SymbolicExpressionTree tree, bool optimizeConstants = true) {
+      double r2;
+
+      // TODO: Initialize constant values randomly
+      // TODO: Restarts
+      if (optimizeConstants) {
         r2 = SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(expressionTreeLinearInterpreter,
           tree,
@@ -100 +120 @@
         }
 
-
       } else {
         var target = problemData.TargetVariableTrainingValues;
+
+        SymbolicRegressionModel model = new SymbolicRegressionModel(
+          problemData.TargetVariable,
+          tree,
+          expressionTreeLinearInterpreter);
+
         var estVals = model.GetEstimatedValues(problemData.Dataset, problemData.TrainingIndices);
         OnlineCalculatorError error;
@@ -109 +134 @@
       }
 
-      var bestR2Result = (DoubleValue)algorithm.Results[BestTrainingQualityResultName].Value;
-      bool better = r2 > bestR2Result.Value;
-      bool equallyGood = r2.IsAlmost(bestR2Result.Value);
-      bool shorter = false;
-
-      if (!better && equallyGood) {
-        shorter = algorithm.BestTrainingSentence != null &&
-          algorithm.Grammar.GetComplexity(algorithm.BestTrainingSentence) > algorithm.Grammar.GetComplexity(symbolString);
-      }
-      if (better || (equallyGood && shorter)) {
-        bestR2Result.Value = r2;
-        algorithm.Results.AddOrUpdateResult(BestTrainingModelResultName, model);
-
-        algorithm.BestTrainingSentence = symbolString;
-      }
+      return r2;
     }
   }