Changeset 8425

Timestamp:

08/07/12 15:14:34 (12 years ago)

Author:

abeham

Message:

#1913: Added several initialization methods (LDA, PCA, and Random)

Location:

branches/NCA

Files:

• . (modified) (1 prop)
• HeuristicLab.Algorithms.NCA/3.3/HeuristicLab.Algorithms.NCA-3.3.csproj (modified) (2 diffs)
• HeuristicLab.Algorithms.NCA/3.3/Initialization (added)
• HeuristicLab.Algorithms.NCA/3.3/Initialization/INCAInitializer.cs (added)
• HeuristicLab.Algorithms.NCA/3.3/Initialization/LDAInitializer.cs (added)
• HeuristicLab.Algorithms.NCA/3.3/Initialization/PCAInitializer.cs (added)
• HeuristicLab.Algorithms.NCA/3.3/Initialization/RandomInitializer.cs (added)
• HeuristicLab.Algorithms.NCA/3.3/NCA.cs (modified) (6 diffs)
• HeuristicLab.Algorithms.NCA/3.3/NeighborhoodComponentsAnalysis.cs (modified) (3 diffs)
• NCA.sln.DotSettings (added)

branches/NCA

@@ -2 +2 @@
 *.user
 TestResults
+_ReSharper.NCA

@@ -2 +2 @@
 *.user
 TestResults
+_ReSharper.NCA

branches/NCA/HeuristicLab.Algorithms.NCA/3.3/HeuristicLab.Algorithms.NCA-3.3.csproj

@@ -89 +89 @@
       <Private>False</Private>
     </Reference>
+    <Reference Include="HeuristicLab.Random-3.3, Version=3.3.0.0, Culture=neutral, PublicKeyToken=ba48961d6f65dcec, processorArchitecture=MSIL" />
     <Reference Include="System" />
     <Reference Include="System.Core" />
@@ -98 +99 @@
   </ItemGroup>
   <ItemGroup>
+    <Compile Include="Initialization\INCAInitializer.cs" />
     <Compile Include="INCAModel.cs" />
+    <Compile Include="Initialization\LDAInitializer.cs" />
+    <Compile Include="Initialization\PCAInitializer.cs" />
+    <Compile Include="Initialization\RandomInitializer.cs" />
     <Compile Include="Matrix.cs" />
     <Compile Include="NCA.cs" />

branches/NCA/HeuristicLab.Algorithms.NCA/3.3/NCA.cs

@@ -20 +20 @@
 #endregion
 
+using System.Linq;
 using HeuristicLab.Algorithms.DataAnalysis;
 using HeuristicLab.Common;
@@ -27 +28 @@
 using HeuristicLab.Parameters;
 using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
+using HeuristicLab.PluginInfrastructure;
 using HeuristicLab.Problems.DataAnalysis;
 
@@ -34 +36 @@
   /// </summary>
   [Item("NCA", "Neihborhood Components Analysis is described in J. Goldberger, S. Roweis, G. Hinton, R. Salakhutdinov. 2005. Neighbourhood Component Analysis. Advances in Neural Information Processing Systems, 17. pp. 513-520.")]
-  [Creatable("Algorithms")]
+  [Creatable("Data Analysis")]
   [StorableClass]
   public sealed class NCA : FixedDataAnalysisAlgorithm<IClassificationProblem> {
@@ -43 +45 @@
     public IValueLookupParameter<IntValue> ReduceDimensionsParameter {
       get { return (IValueLookupParameter<IntValue>)Parameters["ReduceDimensions"]; }
+    }
+    private IConstrainedValueParameter<INCAInitializer> InitializationParameter {
+      get { return (IConstrainedValueParameter<INCAInitializer>)Parameters["Initialization"]; }
     }
     #endregion
@@ -62 +67 @@
       Parameters.Add(new ValueLookupParameter<IntValue>("k", "The k for the nearest neighbor.", new IntValue(1)));
       Parameters.Add(new ValueLookupParameter<IntValue>("ReduceDimensions", "The number of dimensions that NCA should reduce the data to.", new IntValue(2)));
+      Parameters.Add(new ConstrainedValueParameter<INCAInitializer>("Initialization", "Which method should be used to initialize the matrix. Typically LDA (linear discriminant analysis) should provide a good estimate."));
+
+      INCAInitializer defaultInitializer = null;
+      foreach (var initializer in ApplicationManager.Manager.GetInstances<INCAInitializer>().OrderBy(x => x.ItemName)) {
+        if (initializer is LDAInitializer) defaultInitializer = initializer;
+        InitializationParameter.ValidValues.Add(initializer);
+      }
+      if (defaultInitializer != null) InitializationParameter.Value = defaultInitializer;
 
       Problem = new ClassificationProblem();
@@ -75 +88 @@
 
     protected override void Run() {
-      var classification = NeighborhoodComponentsAnalysis.CreateNCASolution(Problem.ProblemData, K.Value, ReduceDimensions.Value);
+      var classification = NeighborhoodComponentsAnalysis.CreateNCASolution(Problem.ProblemData, K.Value, ReduceDimensions.Value, InitializationParameter.Value);
       Results.Add(new Result("ClassificationSolution", "The classification solution.", classification));
     }

branches/NCA/HeuristicLab.Algorithms.NCA/3.3/NeighborhoodComponentsAnalysis.cs

@@ -29 +29 @@
   public class NeighborhoodComponentsAnalysis {
 
-    public static INCAModel Train(IClassificationProblemData data, int k, int reduceDimensions) {
+    public static INCAModel Train(IClassificationProblemData data, int k, int reduceDimensions, INCAInitializer initializer) {
       var instances = data.TrainingIndices.Count();
       var attributes = data.AllowedInputVariables.Count();
 
-      /*alglib.pcabuildbasis(
-    double[,] x,
-    int npoints,
-    int nvars,
-    out int info,
-    out double[] s2,
-    out double[,] v)*/
-
-      var matrix = new double[attributes * reduceDimensions];
-
-      // TODO: make some more clever initialization of matrix
-      var rand = new Random();
-      var counter = 0;
-      for (var i = 0; i < attributes; i++)
-        for (var j = 0; j < reduceDimensions; j++)
-          matrix[counter++] = rand.NextDouble();
+      double[] matrix = initializer.Initialize(data, reduceDimensions);
 
       alglib.mincgstate state;
@@ -60 +45 @@
 
       var transformationMatrix = new double[attributes, reduceDimensions];
-      counter = 0;
+      var counter = 0;
       for (var i = 0; i < attributes; i++)
         for (var j = 0; j < reduceDimensions; j++)
@@ -143 +128 @@
     }
 
-    public static NCAClassificationSolution CreateNCASolution(IClassificationProblemData problemData, int k, int reduceDimensions) {
-      return new NCAClassificationSolution(problemData, Train(problemData, k, reduceDimensions));
+    public static NCAClassificationSolution CreateNCASolution(IClassificationProblemData problemData, int k, int reduceDimensions, INCAInitializer initializer) {
+      return new NCAClassificationSolution(problemData, Train(problemData, k, reduceDimensions, initializer));
     }
 

@@ -2 +2 @@
 *.user
 TestResults
+_ReSharper.NCA