Changeset 17999 for trunk/HeuristicLab.Problems.DataAnalysis

Timestamp:

06/28/21 10:35:19 (3 years ago)

Author:

mkommend

Message:

#3129: Work on variable range calculation of datasets.

Renamed method to calculate variable ranges to a more appropriate method name.
Added possibility to ignore NaNs for range calculation.

Location:

trunk/HeuristicLab.Problems.DataAnalysis/3.4

Files:

• DatasetExtensions.cs (modified) (1 diff)
• Implementation/Regression/RegressionProblemData.cs (modified) (2 diffs)
• Implementation/Regression/ShapeConstrainedRegressionProblemData.cs (modified) (1 diff)

trunk/HeuristicLab.Problems.DataAnalysis/3.4/DatasetExtensions.cs

@@ -96 +96 @@
     }
 
-    public static IntervalCollection GetIntervals(this IDataset dataset) {
-      IntervalCollection intervalCollection = new IntervalCollection();
-      foreach (var variable in dataset.DoubleVariables) { // intervals are only possible for double variables
-        var variableInterval = Interval.GetInterval(dataset.GetDoubleValues(variable));
-        intervalCollection.AddInterval(variable, variableInterval);
+    public static IntervalCollection GetVariableRanges(this IDataset dataset, bool ignoreNaNs = true) {
+      IntervalCollection variableRanges = new IntervalCollection();
+      foreach (var variable in dataset.DoubleVariables) { // ranges can only be calculated for double variables
+        var values = dataset.GetDoubleValues(variable);
+
+        if (ignoreNaNs) {
+          values = values.Where(v => !double.IsNaN(v));
+
+          if (!values.Any()) { //handle values with only NaNs explicitly
+            var emptyInterval = new Interval(double.NaN, double.NaN);
+            variableRanges.AddInterval(variable, emptyInterval);
+            continue;
+          }
+        }
+
+        var interval = Interval.GetInterval(values);
+        variableRanges.AddInterval(variable, interval);
       }
 
-      return intervalCollection;
+      return variableRanges;
     }
 

trunk/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/RegressionProblemData.cs

@@ -130 +130 @@
     private void AfterDeserialization() {
       if (!Parameters.ContainsKey(VariableRangesParameterName)) {
-        var intervalCollection = Dataset.GetIntervals();
+        var variableRanges = Dataset.GetVariableRanges();
         Parameters.Add(new FixedValueParameter<IntervalCollection>(VariableRangesParameterName, intervalCollection));
       }
@@ -164 +164 @@
       Parameters.Add(new ConstrainedValueParameter<StringValue>(TargetVariableParameterName, new ItemSet<StringValue>(variables), variables.Where(x => x.Value == targetVariable).First()));
       if (variableRanges == null) {
-        variableRanges = Dataset.GetIntervals();
+        variableRanges = Dataset.GetVariableRanges();
       }
       Parameters.Add(new FixedValueParameter<IntervalCollection>(VariableRangesParameterName, variableRanges));

trunk/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/ShapeConstrainedRegressionProblemData.cs

@@ -100 +100 @@
         new ShapeConstraint("x", 1, new Interval(0, double.PositiveInfinity), 1.0)
       };
-      defaultVariableRanges = defaultDataset.GetIntervals();
+      defaultVariableRanges = defaultDataset.GetVariableRanges();
       defaultVariableRanges.SetInterval("x", new Interval(0, 6));