Changeset 5894 for trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation

Timestamp:

03/30/11 18:04:03 (14 years ago)

Author:

gkronber

Message:

#1453: Added an ErrorState property to online evaluators to indicate if the result value is valid or if there has been an error in the calculation. Adapted all classes that use one of the online evaluators to check this property.

Location:

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

Files:

• Classification/ClassificationSolution.cs (modified) (1 diff)
• Classification/DiscriminantFunctionClassificationSolution.cs (modified) (1 diff)
• Classification/ThresholdCalculators/NormalDistributionCutPointsThresholdCalculator.cs (modified) (1 diff)
• Regression/RegressionSolution.cs (modified) (1 diff)

trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolution.cs

@@ -85 +85 @@
       IEnumerable<double> originalTestClassValues = ProblemData.Dataset.GetEnumeratedVariableValues(ProblemData.TargetVariable, ProblemData.TestIndizes);
 
-      double trainingAccuracy = OnlineAccuracyEvaluator.Calculate(estimatedTrainingClassValues, originalTrainingClassValues);
-      double testAccuracy = OnlineAccuracyEvaluator.Calculate(estimatedTestClassValues, originalTestClassValues);
+      OnlineEvaluatorError errorState;
+      double trainingAccuracy = OnlineAccuracyEvaluator.Calculate(estimatedTrainingClassValues, originalTrainingClassValues, out errorState);
+      if (errorState != OnlineEvaluatorError.None) trainingAccuracy = double.NaN;
+      double testAccuracy = OnlineAccuracyEvaluator.Calculate(estimatedTestClassValues, originalTestClassValues, out errorState);
+      if (errorState != OnlineEvaluatorError.None) testAccuracy = double.NaN;
 
       TrainingAccuracy = trainingAccuracy;

trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/DiscriminantFunctionClassificationSolution.cs

@@ -105 +105 @@
       IEnumerable<double> originalTestValues = ProblemData.Dataset.GetEnumeratedVariableValues(ProblemData.TargetVariable, ProblemData.TestIndizes);
 
-      double trainingMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues);
-      double testMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTestValues, originalTestValues);
-      double trainingR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues);
-      double testR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTestValues, originalTestValues);
+      OnlineEvaluatorError errorState;
+      double trainingMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues, out errorState);
+      TrainingMeanSquaredError = errorState == OnlineEvaluatorError.None ? trainingMSE : double.NaN;
+      double testMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTestValues, originalTestValues, out errorState);
+      TestMeanSquaredError = errorState == OnlineEvaluatorError.None ? testMSE : double.NaN;
 
-      TrainingMeanSquaredError = trainingMSE;
-      TestMeanSquaredError = testMSE;
-      TrainingRSquared = trainingR2;
-      TestRSquared = testR2;
+      double trainingR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues, out errorState);
+      TrainingRSquared = errorState == OnlineEvaluatorError.None ? trainingR2 : double.NaN;
+      double testR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTestValues, originalTestValues, out errorState);
+      TestRSquared = errorState == OnlineEvaluatorError.None ? testR2 : double.NaN;
     }
 

trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ThresholdCalculators/NormalDistributionCutPointsThresholdCalculator.cs

@@ -64 +64 @@
         double classValue = group.Key;
         double mean, variance;
-        OnlineMeanAndVarianceCalculator.Calculate(estimatedClassValues, out mean, out variance);
-        classMean[classValue] = mean;
-        classStdDev[classValue] = Math.Sqrt(variance);
+        OnlineEvaluatorError meanErrorState, varianceErrorState;
+        OnlineMeanAndVarianceCalculator.Calculate(estimatedClassValues, out mean, out variance, out meanErrorState, out varianceErrorState);
+
+        if (meanErrorState == OnlineEvaluatorError.None && varianceErrorState == OnlineEvaluatorError.None) {
+          classMean[classValue] = mean;
+          classStdDev[classValue] = Math.Sqrt(variance);
+        }
       }
       double[] originalClasses = classMean.Keys.OrderBy(x => x).ToArray();

trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/RegressionSolution.cs

@@ -114 +114 @@
       IEnumerable<double> originalTestValues = ProblemData.Dataset.GetEnumeratedVariableValues(ProblemData.TargetVariable, ProblemData.TestIndizes);
 
-      double trainingMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues);
-      double testMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTestValues, originalTestValues);
-      double trainingR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues);
-      double testR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTestValues, originalTestValues);
-      double trainingRelError = OnlineMeanAbsolutePercentageErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues);
-      double testRelError = OnlineMeanAbsolutePercentageErrorEvaluator.Calculate(estimatedTestValues, originalTestValues);
+      OnlineEvaluatorError errorState;
+      double trainingMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues, out errorState);
+      TrainingMeanSquaredError = errorState == OnlineEvaluatorError.None ? trainingMSE : double.NaN;
+      double testMSE = OnlineMeanSquaredErrorEvaluator.Calculate(estimatedTestValues, originalTestValues, out errorState);
+      TestMeanSquaredError = errorState == OnlineEvaluatorError.None ? testMSE : double.NaN;
 
-      TrainingMeanSquaredError = trainingMSE;
-      TestMeanSquaredError = testMSE;
-      TrainingRSquared = trainingR2;
-      TestRSquared = testR2;
-      TrainingRelativeError = trainingRelError;
-      TestRelativeError = testRelError;
+      double trainingR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues, out errorState);
+      TrainingRSquared = errorState == OnlineEvaluatorError.None ? trainingR2 : double.NaN;
+      double testR2 = OnlinePearsonsRSquaredEvaluator.Calculate(estimatedTestValues, originalTestValues, out errorState);
+      TestRSquared = errorState == OnlineEvaluatorError.None ? testR2 : double.NaN;
+
+      double trainingRelError = OnlineMeanAbsolutePercentageErrorEvaluator.Calculate(estimatedTrainingValues, originalTrainingValues, out errorState);
+      TrainingRelativeError = errorState == OnlineEvaluatorError.None ? trainingRelError : double.NaN;
+      double testRelError = OnlineMeanAbsolutePercentageErrorEvaluator.Calculate(estimatedTestValues, originalTestValues, out errorState);
+      TestRelativeError = errorState == OnlineEvaluatorError.None ? testRelError : double.NaN;
     }