Changeset 136 for trunk/sources/HeuristicLab.StructureIdentification/Evaluation

Timestamp:

04/19/08 15:30:20 (17 years ago)

Author:

gkronber

Message:

fixed a bug in the MSE evaluator which uses an early stopping criterion.
(ticket #29)

File:

• trunk/sources/HeuristicLab.StructureIdentification/Evaluation/EarlyStoppingMeanSquaredErrorEvaluator.cs (modified) (3 diffs)

trunk/sources/HeuristicLab.StructureIdentification/Evaluation/EarlyStoppingMeanSquaredErrorEvaluator.cs

@@ -34 +34 @@
     public override string Description {
       get {
-        return @"Evaluates 'OperatorTree' for samples 'FirstSampleIndex' - 'LastSampleIndex' (inclusive) and calculates the mean-squared-error
+        return @"Evaluates 'OperatorTree' for all samples of the dataset and calculates the mean-squared-error
 for the estimated values vs. the real values of 'TargetVariable'.
 This operator stops the computation as soon as an upper limit for the mean-squared-error is reached.";
@@ -46 +46 @@
 
     public override double Evaluate(IScope scope, IFunction function, int targetVariable, Dataset dataset) {
-      double qualityLimit = GetVariableValue<DoubleData>("QualityLimit", scope, true).Data;
+      double qualityLimit = GetVariableValue<DoubleData>("QualityLimit", scope, false).Data;
       double errorsSquaredSum = 0;
       double targetMean = dataset.GetMean(targetVariable);
@@ -62 +62 @@
         errorsSquaredSum += error * error;
 
-        // check the limit every 10 samples and stop as soon as we hit the limit
-        if(sample % 10 == 9)
-          if(qualityLimit < errorsSquaredSum / dataset.Rows ||
-            double.IsNaN(errorsSquaredSum) ||
-            double.IsInfinity(errorsSquaredSum)) 
-            return errorsSquaredSum / sample; // return estimated MSE (when the remaining errors are on average the same)
+        // check the limit and stop as soon as we hit the limit
+        if(errorsSquaredSum / dataset.Rows >= qualityLimit)
+          return errorsSquaredSum / (sample+1); // return estimated MSE (when the remaining errors are on average the same)
       }
       errorsSquaredSum /= dataset.Rows;