Changeset 2148 for trunk/sources

Timestamp:

07/07/09 13:06:31 (15 years ago)

Author:

mkommend

Message:

ignored dataset rows where the target variable is nan during !SVM learning (ticket #693)

Location:

trunk/sources/HeuristicLab.SupportVectorMachines/3.2

Files:

• SVMHelper.cs (modified) (1 diff)
• SupportVectorEvaluator.cs (modified) (1 diff)

trunk/sources/HeuristicLab.SupportVectorMachines/3.2/SVMHelper.cs

@@ -12 +12 @@
       int rowCount = end - start;
       double[] samples = dataset.Samples;
-      
-      SVM.Node[][] nodes = new SVM.Node[rowCount][];
+
+      List<int> skippedFeatures = new List<int>();
+      for (int i = 0; i < allowedFeatures.Count; i++) {
+        if (dataset.GetRange(allowedFeatures[i].Data, start, end) == 0)
+          skippedFeatures.Add(i);
+      }
+
+      double[] targetVector = new double[rowCount];
+      for (int i = 0; i < rowCount; i++) {
+        double value = samples[(start + i) * dataset.Columns + targetVariable];
+          targetVector[i] = value;
+      }
+      targetVector = targetVector.Where(x=> !double.IsNaN(x)).ToArray();
+
+      SVM.Node[][] nodes = new SVM.Node[targetVector.Length][];
       List<SVM.Node> tempRow;
-      double value;
+      int addedRows = 0;
       for (int row = 0; row < rowCount; row++) {
         tempRow = new List<SVM.Node>();
         for (int col = 0; col < allowedFeatures.Count; col++) {
-          value = samples[(start + row) * dataset.Columns + allowedFeatures[col].Data];
-          if(!double.IsNaN(value))
-            tempRow.Add(new SVM.Node(allowedFeatures[col].Data, value));
+          if (!skippedFeatures.Contains(col)) {
+            double value = samples[(start + row) * dataset.Columns + allowedFeatures[col].Data];
+            if (!double.IsNaN(value))
+              tempRow.Add(new SVM.Node(allowedFeatures[col].Data, value));
+          }
         }
-        nodes[row] = tempRow.ToArray();
+        if (!double.IsNaN(samples[(start + row) * dataset.Columns + targetVariable])) {
+          nodes[addedRows] = tempRow.ToArray();
+          addedRows++;
+        }
       }
 
-      double[] targetVector = new double[rowCount];
-      for (int i = 0; i < rowCount; i++)
-        targetVector[i] = samples[(start + i) * dataset.Columns + targetVariable];
-
-      return new SVM.Problem(rowCount, targetVector, nodes, allowedFeatures.Max(x => x.Data));
+      return new SVM.Problem(targetVector.Length, targetVector, nodes, allowedFeatures.Max(x => x.Data));
     }
   }

trunk/sources/HeuristicLab.SupportVectorMachines/3.2/SupportVectorEvaluator.cs

@@ -56 +56 @@
       SVM.Problem scaledProblem = SVM.Scaling.Scale(problem, modelData.RangeTransform);
 
-      double[,] values = new double[end-start, 2];
-      for (int i = 0; i < end - start; i++) {
+      double[,] values = new double[scaledProblem.Count, 2];
+      for (int i = 0; i < scaledProblem.Count; i++) {
         values[i,0] = SVM.Prediction.Predict(modelData.Model, scaledProblem.X[i]);
         values[i,1] = dataset.GetValue(start + i,targetVariable);