Changeset 2328 for trunk/sources/HeuristicLab.GP.StructureIdentification.Classification

Timestamp:

09/03/09 15:00:23 (15 years ago)

Author:

gkronber

Message:

this is the remaining part of changeset r2327.
Applied changes in modeling plugins that are necessary for the new model analyzer (#722)

• predictor has properties for the lower and upper limit of the predicted value
• added views for predictors that show the limits (also added a new view for GeneticProgrammingModel that shows the size and height of the model)
• Reintroduced TreeEvaluatorInjectors that read a PunishmentFactor and calculate the lower and upper limits for estimated values (limits are set in the tree evaluators)
• Added operators to create Predictors. Changed modeling algorithms to use the predictors for the calculation of final model qualities and variable impacts (to be compatible with the new model analyzer the predictors use a very large PunishmentFactor)
• replaced all private implementations of double.IsAlmost and use HL.Commons instead (see #733 r2324)
• Implemented operator SolutionExtractor and moved BestSolutionStorer from HL.Logging to HL.Modeling (fixes #734)

Location:

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3

Files:

• ClassificationMeanSquaredErrorEvaluator.cs (modified) (3 diffs)
• HeuristicLab.GP.StructureIdentification.Classification-3.3.csproj (modified) (1 diff)
• HeuristicLabGPClassificationPlugin.cs (modified) (1 diff)
• MulticlassModeller.cs (modified) (6 diffs)
• MulticlassOneVsOneAnalyzer.cs (modified) (6 diffs)

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3/ClassificationMeanSquaredErrorEvaluator.cs

@@ -23 +23 @@
 using HeuristicLab.Core;
 using HeuristicLab.Data;
+using HeuristicLab.Common;
 
 namespace HeuristicLab.GP.StructureIdentification.Classification {
@@ -49 +50 @@
           // on the lower end and upper end only add linear error if the absolute error is larger than 1
           // the error>1.0 constraint is needed for balance because in the interval ]-1, 1[ the squared error is smaller than the absolute error
-          if ((IsEqual(original, classes[0]) && error < -1.0) ||
-            (IsEqual(original, classes[classes.Length - 1]) && error > 1.0)) {
+          if ((original.IsAlmost(classes[0]) && error < -1.0) ||
+            (original.IsAlmost(classes[classes.Length - 1]) && error > 1.0)) {
             errorsSquaredSum += Math.Abs(error); // only add linear error below the smallest class or above the largest class
           } else {
@@ -71 +72 @@
       mse.Data = errorsSquaredSum;
     }
-
-    private bool IsEqual(double x, double y) {
-      return Math.Abs(x - y) < EPSILON;
-    }
   }
 }

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3/HeuristicLab.GP.StructureIdentification.Classification-3.3.csproj

@@ -94 +94 @@
   </ItemGroup>
   <ItemGroup>
+    <ProjectReference Include="..\..\HeuristicLab.Common\3.2\HeuristicLab.Common-3.2.csproj">
+      <Project>{1FC004FC-59AF-4249-B1B6-FF25873A20E4}</Project>
+      <Name>HeuristicLab.Common-3.2</Name>
+    </ProjectReference>
     <ProjectReference Include="..\..\HeuristicLab.Core\3.2\HeuristicLab.Core-3.2.csproj">
       <Project>{F43B59AB-2B8C-4570-BC1E-15592086517C}</Project>

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3/HeuristicLabGPClassificationPlugin.cs

@@ -28 +28 @@
   [ClassInfo(Name = "HeuristicLab.GP.StructureIdentification.Classification-3.3")]
   [PluginFile(Filename = "HeuristicLab.GP.StructureIdentification.Classification-3.3.dll", Filetype = PluginFileType.Assembly)]
+  [Dependency(Dependency = "HeuristicLab.Common-3.2")]
   [Dependency(Dependency = "HeuristicLab.Core-3.2")]
   [Dependency(Dependency = "HeuristicLab.Data-3.2")]

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3/MulticlassModeller.cs

@@ -22 +22 @@
 using System;
 using System.Collections.Generic;
+using HeuristicLab.Common;
 using HeuristicLab.Core;
 using HeuristicLab.Data;
@@ -83 +84 @@
             double[] row = new double[dataset.Columns];
             double targetValue = origDataset.GetValue(k, targetVariable);
-            if (IsEqual(targetValue, classAValue)) {
+            if (targetValue.IsAlmost(classAValue)) {
               for (int l = 0; l < row.Length; l++) {
                 row[l] = origDataset.GetValue(k, l);
@@ -89 +90 @@
               row[targetVariable] = 0;
               rows.Add(row);
-            } else if (IsEqual(targetValue, classBValue)) {
+            } else if (targetValue.IsAlmost(classBValue)) {
               for (int l = 0; l < row.Length; l++) {
                 row[l] = origDataset.GetValue(k, l);
@@ -102 +103 @@
             double[] row = new double[dataset.Columns];
             double targetValue = origDataset.GetValue(k, targetVariable);
-            if (IsEqual(targetValue, classAValue)) {
+            if (targetValue.IsAlmost(classAValue)) {
               for (int l = 0; l < row.Length; l++) {
                 row[l] = origDataset.GetValue(k, l);
@@ -108 +109 @@
               row[targetVariable] = 0;
               rows.Add(row);
-            } else if (IsEqual(targetValue, classBValue)) {
+            } else if (targetValue.IsAlmost(classBValue)) {
               for (int l = 0; l < row.Length; l++) {
                 row[l] = origDataset.GetValue(k, l);
@@ -141 +142 @@
       return null;
     }
-
-    private bool IsEqual(double x, double y) {
-      return Math.Abs(x - y) < EPSILON;
-    }
   }
 }

trunk/sources/HeuristicLab.GP.StructureIdentification.Classification/3.3/MulticlassOneVsOneAnalyzer.cs

@@ -25 +25 @@
 using HeuristicLab.DataAnalysis;
 using HeuristicLab.GP.Interfaces;
+using HeuristicLab.Common;
 
 namespace HeuristicLab.GP.StructureIdentification.Classification {
@@ -44 +45 @@
     private const string TREEEVALUATOR = "TreeEvaluator";
 
-    private const double EPSILON = 1E-6;
     public override string Description {
       get { return @"TASK"; }
@@ -56 +56 @@
       AddVariableInfo(new VariableInfo(CLASSAVALUE, "The original class value of the class A in the subscope", typeof(DoubleData), VariableKind.In));
       AddVariableInfo(new VariableInfo(CLASSBVALUE, "The original class value of the class B in the subscope", typeof(DoubleData), VariableKind.In));
-      AddVariableInfo(new VariableInfo(TRAININGSAMPLESSTART, "The start of training samples in the original dataset", typeof(IntData), VariableKind.In));
-      AddVariableInfo(new VariableInfo(TRAININGSAMPLESEND, "The end of training samples in the original dataset", typeof(IntData), VariableKind.In));
       AddVariableInfo(new VariableInfo(SAMPLESSTART, "The start of samples in the original dataset", typeof(IntData), VariableKind.In));
       AddVariableInfo(new VariableInfo(SAMPLESEND, "The end of samples in the original dataset", typeof(IntData), VariableKind.In));
@@ -84 +82 @@
 
         ITreeEvaluator evaluator = GetVariableValue<ITreeEvaluator>(TREEEVALUATOR, bestScope, true);
-        evaluator.PrepareForEvaluation(dataset, targetVariable, trainingSamplesStart, trainingSamplesEnd, gpModel.FunctionTree);
+        evaluator.PrepareForEvaluation(dataset, gpModel.FunctionTree);
         for(int i = 0; i < (samplesEnd - samplesStart); i++) {
           double est = evaluator.Evaluate(i + samplesStart);
@@ -110 +108 @@
           }
         }
-        if(IsEqual(originalClassValue, estimatedClassValue) && sameVotes == 0) correctlyClassified++;
+        if(originalClassValue.IsAlmost(estimatedClassValue) && sameVotes == 0) correctlyClassified++;
       }
 
@@ -122 +120 @@
     private void CastVote(int[,] votes, int sample, double votedClass, ItemList<DoubleData> classValues) {
       for(int i = 0; i < classValues.Count; i++) {
-        if(IsEqual(classValues[i].Data, votedClass)) votes[sample, i]++;
+        if(classValues[i].Data.IsAlmost(votedClass)) votes[sample, i]++;
       }
-    }
-
-    private bool IsEqual(double x, double y) {
-      return Math.Abs(x - y) < EPSILON;
     }
   }