Changeset 17457 for branches/2521_ProblemRefactoring/HeuristicLab.Problems.DataAnalysis

Timestamp:

02/28/20 10:56:16 (5 years ago)

Author:

mkommend

Message:

#2521: Merged trunk changes from 15684-HEAD into the branch.

Location:

branches/2521_ProblemRefactoring

Files:

• . (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis/3.4 (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolutionVariableImpactsCalculator.cs (modified) (8 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/DiscriminantFunctionClassificationModel.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Interval.cs (modified) (6 diffs)

branches/2521_ProblemRefactoring/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolutionVariableImpactsCalculator.cs

@@ -26 +26 @@
 using System.Collections.Generic;
 using System.Linq;
+using HEAL.Attic;
 using HeuristicLab.Common;
 using HeuristicLab.Core;
 using HeuristicLab.Data;
 using HeuristicLab.Parameters;
-using HEAL.Attic;
 using HeuristicLab.Random;
 
@@ -173 +173 @@
 
       IList originalValues = null;
-      IList replacementValues = GetReplacementValues(modifiableDataset, variableName, model, rows, targetValues, out originalValues, replacementMethod, factorReplacementMethod);
-
-      double newValue = CalculateQualityForReplacement(model, modifiableDataset, variableName, originalValues, rows, replacementValues, targetValues);
+      IList replacementValues = GetReplacementValues(modifiableDataset, variableName, model, problemData.AllowedInputVariables, rows, targetValues, out originalValues, replacementMethod, factorReplacementMethod);
+
+      double newValue = CalculateQualityForReplacement(model, modifiableDataset, problemData.AllowedInputVariables, variableName, originalValues, rows, replacementValues, targetValues);
       double impact = quality - newValue;
 
@@ -184 +184 @@
       string variableName,
       IClassificationModel model,
+      IEnumerable<string> allowedInputVariables,
       IEnumerable<int> rows,
       IEnumerable<double> targetValues,
@@ -196 +197 @@
       } else if (modifiableDataset.VariableHasType<string>(variableName)) {
         originalValues = modifiableDataset.GetReadOnlyStringValues(variableName).ToList();
-        replacementValues = GetReplacementValuesForString(model, modifiableDataset, variableName, rows, (List<string>)originalValues, targetValues, factorReplacementMethod);
+        replacementValues = GetReplacementValuesForString(model, modifiableDataset, allowedInputVariables, variableName, rows, (List<string>)originalValues, targetValues, factorReplacementMethod);
       } else {
         throw new NotSupportedException("Variable not supported");
@@ -254 +255 @@
     private static IList GetReplacementValuesForString(IClassificationModel model,
       ModifiableDataset modifiableDataset,
+      IEnumerable<string> allowedInputVariables,
       string variableName,
       IEnumerable<int> rows,
@@ -270 +272 @@
             List<string> curReplacementValues = Enumerable.Repeat(repl, modifiableDataset.Rows).ToList();
             //fholzing: this result could be used later on (theoretically), but is neglected for better readability/method consistency 
-            var newValue = CalculateQualityForReplacement(model, modifiableDataset, variableName, originalValues, rows, curReplacementValues, targetValues);
+            var newValue = CalculateQualityForReplacement(model, modifiableDataset, allowedInputVariables, variableName, originalValues, rows, curReplacementValues, targetValues);
             var curQuality = newValue;
 
@@ -308 +310 @@
       IClassificationModel model,
       ModifiableDataset modifiableDataset,
+      IEnumerable<string> allowedInputVariables,
       string variableName,
       IList originalValues,
@@ -317 +320 @@
       var discModel = model as IDiscriminantFunctionClassificationModel;
       if (discModel != null) {
-        var problemData = new ClassificationProblemData(modifiableDataset, modifiableDataset.VariableNames, model.TargetVariable);
+        var problemData = new ClassificationProblemData(modifiableDataset, allowedInputVariables, model.TargetVariable);
         discModel.RecalculateModelParameters(problemData, rows);
       }

branches/2521_ProblemRefactoring/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/DiscriminantFunctionClassificationModel.cs

@@ -74 +74 @@
       classValues = (double[])original.classValues.Clone();
       thresholds = (double[])original.thresholds.Clone();
+      thresholdCalculator = (IDiscriminantFunctionThresholdCalculator)original.thresholdCalculator.Clone();
     }
 

branches/2521_ProblemRefactoring/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Interval.cs

@@ -37 +37 @@
     protected Interval(StorableConstructorFlag _) { }
 
+    /// <summary>
+    /// Creates an interval with given bounds, where lower bound must be smaller than
+    /// the upper bound. Floating point precision errors trough calculations are fixed by,
+    /// checking if the intervals are almost the same (E-12). If this is the case, the bounds 
+    /// will be set to the bound closer to zero.
+    /// </summary>
+    /// <param name="lowerBound">Lower bound of the interval</param>
+    /// <param name="upperBound">Upper bound of the interval</param>
     public Interval(double lowerBound, double upperBound) {
+      if (lowerBound.IsAlmost(upperBound)) {
+        //If the bounds go over zero
+        if (lowerBound <= 0 && upperBound >= 0) {
+          lowerBound = 0.0;
+          upperBound = 0.0;
+          //Interval is negative
+        } else if (upperBound < 0) {
+          lowerBound = upperBound;
+          //Interval is positive
+        } else {
+          upperBound = lowerBound;
+        }
+      }
+
       if (lowerBound > upperBound)
         throw new ArgumentException("LowerBound must be smaller than UpperBound.");
@@ -58 +80 @@
                 double.IsNaN(LowerBound) || double.IsNaN(UpperBound);
       }
+    }
+
+    /// <summary>
+    /// True if the interval is positive without zero
+    /// </summary>
+    public bool IsPositive {
+      get => LowerBound > 0.0; 
+    }
+
+    /// <summary>
+    /// True if the interval is negative without zero
+    /// </summary>
+    public bool IsNegative {
+      get => UpperBound < 0.0;
     }
 
@@ -128 +164 @@
     }
 
-    //mkommend: Division by intervals containing 0 is implemented as defined in
+    //Division by intervals containing 0 is implemented as defined in
     //http://en.wikipedia.org/wiki/Interval_arithmetic
     public static Interval Divide(Interval a, Interval b) {
@@ -171 +207 @@
     public static Interval Tangens(Interval a) {
       return Interval.Divide(Interval.Sine(a), Interval.Cosine(a));
-    }  
+    }
     public static Interval HyperbolicTangent(Interval a) {
       return new Interval(Math.Tanh(a.LowerBound), Math.Tanh(a.UpperBound));
@@ -206 +242 @@
       if (a.UpperBound <= 0) return new Interval(a.UpperBound * a.UpperBound, a.LowerBound * a.LowerBound);     // interval is negative
       else if (a.LowerBound >= 0) return new Interval(a.LowerBound * a.LowerBound, a.UpperBound * a.UpperBound); // interval is positive
-      else return new Interval(0, Math.Max(a.LowerBound*a.LowerBound, a.UpperBound*a.UpperBound)); // interval goes over zero
+      else return new Interval(0, Math.Max(a.LowerBound * a.LowerBound, a.UpperBound * a.UpperBound)); // interval goes over zero
     }
 
@@ -235 +271 @@
       var absLower = Math.Abs(a.LowerBound);
       var absUpper = Math.Abs(a.UpperBound);
-      return new Interval(Math.Min(absLower, absUpper), Math.Max(absLower, absUpper));
+      var min = Math.Min(absLower, absUpper);
+      var max = Math.Max(absLower, absUpper);
+
+      if (a.Contains(0.0)) {
+        min = 0.0;
+      }
+
+      return new Interval(min, max);
+    }
+
+    public static Interval AnalyticalQuotient(Interval a, Interval b) {
+      var dividend = a;
+      var divisor = Add(Square(b), new Interval(1.0, 1.0));
+      divisor = SquareRoot(divisor);
+
+      var quotient = Divide(dividend, divisor);
+      return quotient;
     }
     #endregion