Changeset 11361 for trunk

Timestamp:

09/14/14 13:30:30 (10 years ago)

Author:

bburlacu

Message:

#2234: Added the option to shuffle the crossvalidation folds (this option is on by default since libsvm does it too). Implemented stratified fold generation for classification data (ensures similar label distribution in each fold).

File:

• trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/SupportVectorMachine/SupportVectorMachineUtil.cs (modified) (7 diffs)

trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/SupportVectorMachine/SupportVectorMachineUtil.cs

@@ -26 +26 @@
 using System.Threading.Tasks;
 using HeuristicLab.Common;
+using HeuristicLab.Core;
 using HeuristicLab.Data;
 using HeuristicLab.Problems.DataAnalysis;
+using HeuristicLab.Random;
 using LibSVM;
 
@@ -52 +54 @@
           // => don't add NaN values in the dataset to the sparse SVM matrix representation
           if (!double.IsNaN(value)) {
-            tempRow.Add(new svm_node() { index = colIndex, value = value }); // nodes must be sorted in ascending ordered by column index
+            tempRow.Add(new svm_node() { index = colIndex, value = value });
+            // nodes must be sorted in ascending ordered by column index
             if (colIndex > maxNodeIndex) maxNodeIndex = colIndex;
           }
@@ -84 +87 @@
     }
 
-    public static void CrossValidate(IDataAnalysisProblemData problemData, svm_parameter parameters, int numberOfFolds, out double avgTestMse) {
-      var partitions = GenerateSvmPartitions(problemData, numberOfFolds);
-      CalculateCrossValidationPartitions(partitions, parameters, out avgTestMse);
-    }
-
-    public static svm_parameter GridSearch(IDataAnalysisProblemData problemData, int numberOfFolds, Dictionary<string, IEnumerable<double>> parameterRanges, int maxDegreeOfParallelism = 1) {
+    public static double CrossValidate(IDataAnalysisProblemData problemData, svm_parameter parameters, int numberOfFolds, bool shuffleFolds = true) {
+      var partitions = GenerateSvmPartitions(problemData, numberOfFolds, shuffleFolds);
+      return CalculateCrossValidationPartitions(partitions, parameters);
+    }
+
+    public static svm_parameter GridSearch(IDataAnalysisProblemData problemData, Dictionary<string, IEnumerable<double>> parameterRanges, int numberOfFolds, bool shuffleFolds = true, int maxDegreeOfParallelism = 1) {
       DoubleValue mse = new DoubleValue(Double.MaxValue);
       var bestParam = DefaultParameters();
       var crossProduct = parameterRanges.Values.CartesianProduct();
       var setters = parameterRanges.Keys.Select(GenerateSetter).ToList();
-      var partitions = GenerateSvmPartitions(problemData, numberOfFolds);
-      Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism }, parameterCombination => {
+      var partitions = GenerateSvmPartitions(problemData, numberOfFolds, shuffleFolds);
+      Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism },
+      parameterCombination => {
         var parameters = DefaultParameters();
         var parameterValues = parameterCombination.ToList();
-        for (int i = 0; i < parameterValues.Count; ++i) {
+        for (int i = 0; i < parameterValues.Count; ++i)
           setters[i](parameters, parameterValues[i]);
-        }
-        double testMse;
-        CalculateCrossValidationPartitions(partitions, parameters, out testMse);
+
+        double testMse = CalculateCrossValidationPartitions(partitions, parameters);
         if (testMse < mse.Value) {
           lock (mse) {
@@ -113 +116 @@
     }
 
-    private static void CalculateCrossValidationPartitions(Tuple<svm_problem, svm_problem>[] partitions, svm_parameter parameters, out double avgTestMse) {
-      avgTestMse = 0;
+    private static double CalculateCrossValidationPartitions(Tuple<svm_problem, svm_problem>[] partitions, svm_parameter parameters) {
+      double avgTestMse = 0;
       var calc = new OnlineMeanSquaredErrorCalculator();
       foreach (Tuple<svm_problem, svm_problem> tuple in partitions) {
@@ -126 +129 @@
       }
       avgTestMse /= partitions.Length;
-    }
-
-
-    private static Tuple<svm_problem, svm_problem>[] GenerateSvmPartitions(IDataAnalysisProblemData problemData, int numberOfFolds) {
-      var folds = GenerateFolds(problemData, numberOfFolds).ToList();
+      return avgTestMse;
+    }
+
+    private static Tuple<svm_problem, svm_problem>[] GenerateSvmPartitions(IDataAnalysisProblemData problemData, int numberOfFolds, bool shuffleFolds = true) {
+      var folds = GenerateFolds(problemData, numberOfFolds, shuffleFolds).ToList();
       var targetVariable = GetTargetVariableName(problemData);
       var partitions = new Tuple<svm_problem, svm_problem>[numberOfFolds];
@@ -144 +147 @@
     }
 
+    public static IEnumerable<IEnumerable<int>> GenerateFolds(IDataAnalysisProblemData problemData, int numberOfFolds, bool shuffleFolds = true) {
+      var random = new MersenneTwister((uint)Environment.TickCount);
+      if (problemData is IRegressionProblemData) {
+        var trainingIndices = shuffleFolds ? problemData.TrainingIndices.OrderBy(x => random.Next()) : problemData.TrainingIndices;
+        return GenerateFolds(trainingIndices, problemData.TrainingPartition.Size, numberOfFolds);
+      }
+      if (problemData is IClassificationProblemData) {
+        // when shuffle is enabled do stratified folds generation, some folds may have zero elements 
+        // otherwise, generate folds normally
+        return shuffleFolds ? GenerateFoldsStratified(problemData as IClassificationProblemData, numberOfFolds, random) : GenerateFolds(problemData.TrainingIndices, problemData.TrainingPartition.Size, numberOfFolds);
+      }
+      throw new ArgumentException("Problem data is neither regression or classification problem data.");
+    }
+
     /// <summary>
-    /// Generate a collection of row sequences corresponding to folds in the data (used for crossvalidation)
+    /// Stratified fold generation from classification data. Stratification means that we ensure the same distribution of class labels for each fold.
+    /// The samples are grouped by class label and each group is split into @numberOfFolds parts. The final folds are formed from the joining of 
+    /// the corresponding parts from each class label.
     /// </summary>
-    /// <remarks>This method is aimed to be lightweight and as such does not clone the dataset.</remarks>
-    /// <param name="problemData">The problem data</param>
-    /// <param name="numberOfFolds">The number of folds to generate</param>
-    /// <returns>A sequence of folds representing each a sequence of row numbers</returns>
-    private static IEnumerable<IEnumerable<int>> GenerateFolds(IDataAnalysisProblemData problemData, int numberOfFolds) {
-      int size = problemData.TrainingPartition.Size;
-      int f = size / numberOfFolds, r = size % numberOfFolds; // number of folds rounded to integer and remainder
-      int start = 0, end = f;
-      for (int i = 0; i < numberOfFolds; ++i) {
-        if (r > 0) { ++end; --r; }
-        yield return problemData.TrainingIndices.Skip(start).Take(end - start);
-        start = end;
-        end += f;
+    /// <param name="problemData">The classification problem data.</param>
+    /// <param name="numberOfFolds">The number of folds in which to split the data.</param>
+    /// <param name="random">The random generator used to shuffle the folds.</param>
+    /// <returns>An enumerable sequece of folds, where a fold is represented by a sequence of row indices.</returns>
+    private static IEnumerable<IEnumerable<int>> GenerateFoldsStratified(IClassificationProblemData problemData, int numberOfFolds, IRandom random) {
+      var values = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices);
+      var valuesIndices = problemData.TrainingIndices.Zip(values, (i, v) => new { Index = i, Value = v }).ToList();
+      IEnumerable<IEnumerable<IEnumerable<int>>> foldsByClass = valuesIndices.GroupBy(x => x.Value, x => x.Index).Select(g => GenerateFolds(g, g.Count(), numberOfFolds));
+      var enumerators = foldsByClass.Select(f => f.GetEnumerator()).ToList();
+      while (enumerators.All(e => e.MoveNext())) {
+        yield return enumerators.SelectMany(e => e.Current).OrderBy(x => random.Next()).ToList();
+      }
+    }
+
+    private static IEnumerable<IEnumerable<T>> GenerateFolds<T>(IEnumerable<T> values, int valuesCount, int numberOfFolds) {
+      // if number of folds is greater than the number of values, some empty folds will be returned
+      if (valuesCount < numberOfFolds) {
+        for (int i = 0; i < numberOfFolds; ++i)
+          yield return i < valuesCount ? values.Skip(i).Take(1) : Enumerable.Empty<T>();
+      } else {
+        int f = valuesCount / numberOfFolds, r = valuesCount % numberOfFolds; // number of folds rounded to integer and remainder
+        int start = 0, end = f;
+        for (int i = 0; i < numberOfFolds; ++i) {
+          if (r > 0) {
+            ++end;
+            --r;
+          }
+          yield return values.Skip(start).Take(end - start);
+          start = end;
+          end += f;
+        }
       }
     }
@@ -183 +220 @@
       throw new ArgumentException("Problem data is neither regression or classification problem data.");
     }
-
   }
 }