Changeset 2550

Timestamp:

12/10/09 15:57:14 (15 years ago)

Author:

gkronber

Message:

Implemented #812 (Static methods for SVM operators).

Location:

trunk/sources/HeuristicLab.SupportVectorMachines/3.2

Files:

• PredictorBuilder.cs (modified) (1 diff)
• SupportVectorCreator.cs (modified) (3 diffs)

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

@@ -61 +61 @@
       double punishmentFactor = GetVariableValue<DoubleData>("PunishmentFactor", scope, true).Data;
 
-      
+
       ItemList inputVariables = GetVariableValue<ItemList>("InputVariables", scope, true);
       var inputVariableNames = from x in inputVariables
                                select ((StringData)x).Data;
 
-      double mean = ds.GetMean(targetVariable, start, end);
-      double range = ds.GetRange(targetVariable, start, end);
-
-      Predictor predictor = new Predictor(model, targetVariable, inputVariableNames, minTimeOffset, maxTimeOffset);
-      predictor.LowerPredictionLimit = mean - punishmentFactor * range;
-      predictor.UpperPredictionLimit = mean + punishmentFactor * range;
+      var predictor = CreatePredictor(model, ds, targetVariable, inputVariableNames, punishmentFactor, start, end, minTimeOffset, maxTimeOffset);
       scope.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("Predictor"), predictor));
       return null;
     }
+
+    public static Predictor CreatePredictor(SVMModel model, Dataset ds, string targetVariable, IEnumerable<string> inputVariables, double punishmentFactor,
+      int start, int end, int minTimeOffset, int maxTimeOffset) {
+      Predictor predictor = new Predictor(model, targetVariable, inputVariables, minTimeOffset, maxTimeOffset);
+      double mean = ds.GetMean(targetVariable, start, end);
+      double range = ds.GetRange(targetVariable, start, end);
+      predictor.LowerPredictionLimit = mean - punishmentFactor * range;
+      predictor.UpperPredictionLimit = mean + punishmentFactor * range;
+      return predictor;
+    }
   }
 }

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

@@ -71 +71 @@
       Dataset dataset = GetVariableValue<Dataset>("Dataset", scope, true);
       string targetVariable = GetVariableValue<StringData>("TargetVariable", scope, true).Data;
-      int targetVariableIndex = dataset.GetVariableIndex(targetVariable);
       ItemList inputVariables = GetVariableValue<ItemList>("InputVariables", scope, true);
       var inputVariableNames = from x in inputVariables
@@ -84 +83 @@
       string svmKernelType = GetVariableValue<StringData>("SVMKernelType", scope, true).Data;
 
-      //extract SVM parameters from scope and set them
-      SVM.Parameter parameter = new SVM.Parameter();
-      parameter.SvmType = (SVM.SvmType)Enum.Parse(typeof(SVM.SvmType), svmType, true);
-      parameter.KernelType = (SVM.KernelType)Enum.Parse(typeof(SVM.KernelType), svmKernelType, true);
-      parameter.C = GetVariableValue<DoubleData>("SVMCost", scope, true).Data;
-      parameter.Nu = GetVariableValue<DoubleData>("SVMNu", scope, true).Data;
-      parameter.Gamma = GetVariableValue<DoubleData>("SVMGamma", scope, true).Data;
-      parameter.CacheSize = 500;
-      parameter.Probability = false;
+      double cost = GetVariableValue<DoubleData>("SVMCost", scope, true).Data;
+      double nu = GetVariableValue<DoubleData>("SVMNu", scope, true).Data;
+      double gamma = GetVariableValue<DoubleData>("SVMGamma", scope, true).Data;
 
-      SVM.Problem problem = SVMHelper.CreateSVMProblem(dataset, targetVariableIndex, inputVariableNames, start, end, minTimeOffset, maxTimeOffset);
-      SVM.RangeTransform rangeTransform = SVM.RangeTransform.Compute(problem);
-      SVM.Problem scaledProblem = rangeTransform.Scale(problem);
-
-      SVM.Model model = StartTraining(scaledProblem, parameter);
-      if (!abortRequested) {
-        //persist variables in scope
-        SVMModel modelData = new SVMModel();
-        modelData.Model = model;
-        modelData.RangeTransform = rangeTransform;
-        scope.AddVariable(new Variable(scope.TranslateName("SVMModel"), modelData));
-        return null;
-      } else {
-        return new AtomicOperation(this, scope);
-      }
-    }
-
-    private SVM.Model StartTraining(SVM.Problem scaledProblem, SVM.Parameter parameter) {
-      SVM.Model model = null;
+      SVMModel modelData = null;
       lock (locker) {
         if (!abortRequested) {
           trainingThread = new Thread(() => {
-            model = SVM.Training.Train(scaledProblem, parameter);
+            modelData = TrainModel(dataset, targetVariable, inputVariableNames,
+                                   start, end, minTimeOffset, maxTimeOffset,
+                                   svmType, svmKernelType,
+                                   cost, nu, gamma);
           });
           trainingThread.Start();
@@ -125 +103 @@
         trainingThread = null;
       }
+
+
+      if (!abortRequested) {
+        //persist variables in scope
+        scope.AddVariable(new Variable(scope.TranslateName("SVMModel"), modelData));
+        return null;
+      } else {
+        return new AtomicOperation(this, scope);
+      }
+    }
+
+    public static SVMModel TrainRegressionModel(
+      Dataset dataset, string targetVariable, IEnumerable<string> inputVariables,
+      int start, int end,
+      double cost, double nu, double gamma) {
+      return TrainModel(dataset, targetVariable, inputVariables, start, end, 0, 0, "NU_SVR", "RBF", cost, nu, gamma);
+    }
+
+    public static SVMModel TrainModel(
+      Dataset dataset, string targetVariable, IEnumerable<string> inputVariables,
+      int start, int end,
+      int minTimeOffset, int maxTimeOffset,
+      string svmType, string kernelType,
+      double cost, double nu, double gamma) {
+      int targetVariableIndex = dataset.GetVariableIndex(targetVariable);
+
+      //extract SVM parameters from scope and set them
+      SVM.Parameter parameter = new SVM.Parameter();
+      parameter.SvmType = (SVM.SvmType)Enum.Parse(typeof(SVM.SvmType), svmType, true);
+      parameter.KernelType = (SVM.KernelType)Enum.Parse(typeof(SVM.KernelType), kernelType, true);
+      parameter.C = cost;
+      parameter.Nu = nu;
+      parameter.Gamma = gamma;
+      parameter.CacheSize = 500;
+      parameter.Probability = false;
+
+
+      SVM.Problem problem = SVMHelper.CreateSVMProblem(dataset, targetVariableIndex, inputVariables, start, end, minTimeOffset, maxTimeOffset);
+      SVM.RangeTransform rangeTransform = SVM.RangeTransform.Compute(problem);
+      SVM.Problem scaledProblem = rangeTransform.Scale(problem);
+      var model = new SVMModel();
+
+      model.Model = SVM.Training.Train(scaledProblem, parameter);
+      model.RangeTransform = rangeTransform;
+
       return model;
     }