Changeset 10482

Timestamp:

02/20/14 14:47:17 (11 years ago)

Author:

gkronber

Message:

#2093 added a notice that the log residual evaluator does not apply linear scaling

File:

• branches/LogResidualEvaluator/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SingleObjective/Evaluators/SymbolicRegressionLogResidualEvaluator.cs (modified) (4 diffs)

branches/LogResidualEvaluator/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SingleObjective/Evaluators/SymbolicRegressionLogResidualEvaluator.cs

@@ -30 +30 @@
 
 namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
-  [Item("Log Residual Evaluator", "Evaluator for symbolic regression models that calculates the mean of logarithmic absolute residuals avg(log( 1 + abs(y' - y)))." +
+  [Item("Log Residual Evaluator", "Evaluator for symbolic regression models that calculates the mean of logarithmic absolute residuals avg(log( 1 + abs(y' - y)))" + 
+                                  "This evaluator does not perform linear scaling!" +
                                   "This evaluator can be useful if the modeled function contains discontinuities (e.g. 1/x). " +
-                                  "In preliminary experiments, we found that this improves success rates for data sets with a few very large/small target values " +
-                                  "(e.g. Korns benchmark instances containing inverses of near zero values). In these cases the squared error or absolute " +
-                                  "error puts too much emphasis on modeling the outlier values. Using log-residuals instead of e.g. squared residuals has the " +
-                                  "effect that smaller residuals have a stronger impact on the total quality compared to the very large residuals occuring with such extreme values." +
+                                  "For some data sets (e.g. Korns benchmark instances containing inverses of near zero values) the squared error or absolute " +
+                                  "error put too much emphasis on modeling the outlier values. Using log-residuals instead has the " +
+                                  "effect that smaller residuals have a stronger impact on the total quality compared to the large residuals." +
                                   "This effects GP convergence because functional fragments which are necessary to explain small variations are also more likely" +
                                   " to stay in the population. This is useful even when the actual objective function is mean of squared errors.")]
@@ -57 +57 @@
       IEnumerable<int> rows = GenerateRowsToEvaluate();
 
-      double quality = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, solution, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, ProblemDataParameter.ActualValue, rows, ApplyLinearScalingParameter.ActualValue.Value);
+      double quality = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, solution, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, ProblemDataParameter.ActualValue, rows);
       QualityParameter.ActualValue = new DoubleValue(quality);
 
@@ -63 +63 @@
     }
 
-    public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
+    public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IRegressionProblemData problemData, IEnumerable<int> rows) {
       IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
       IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
@@ -82 +82 @@
       SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = context;
       EstimationLimitsParameter.ExecutionContext = context;
-      ApplyLinearScalingParameter.ExecutionContext = context;
 
-      double mlr = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, tree, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, problemData, rows, ApplyLinearScalingParameter.ActualValue.Value);
+      double mlr = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, tree, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, problemData, rows);
 
       SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = null;
       EstimationLimitsParameter.ExecutionContext = null;
-      ApplyLinearScalingParameter.ExecutionContext = null;
 
       return mlr;