Changeset 16461 for branches/2974_Constants_Optimization/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SingleObjective/Evaluators

Timestamp:

12/28/18 15:44:31 (6 years ago)

Author:

mkommend

Message:

#2974: Added unit tests and refactoring.

File:

• branches/2974_Constants_Optimization/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SingleObjective/Evaluators/ConstantsOptimizationEvaluator.cs (modified) (6 diffs)

branches/2974_Constants_Optimization/HeuristicLab.Problems.DataAnalysis.Symbolic.Regression/3.4/SingleObjective/Evaluators/ConstantsOptimizationEvaluator.cs

@@ -131 +131 @@
     }
 
-
     private static readonly object locker = new object();
     public override IOperation InstrumentedApply() {
@@ -189 +188 @@
     }
 
+    [ThreadStatic]
+    private static double[,] x = null;
+    [ThreadStatic]
+    private static IDataset ds = null;
+    [ThreadStatic]
+    private static Dictionary<DataForVariable, AutoDiff.Variable> parameters;
+
     public static double OptimizeConstants(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter,
       ISymbolicExpressionTree tree, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling,
@@ -195 +201 @@
       bool updateConstantsInTree = true, Action<double[], double, object> iterationCallback = null, EvaluationsCounter counter = null) {
 
+      if (ds == null || ds != problemData.Dataset) {
+        ds = problemData.Dataset;
+        parameters = ConstantsOptimization.Util.ExtractParameters(problemData.Dataset);
+        x = ConstantsOptimization.Util.ExtractData(ds, parameters.Keys, rows);
+        
+      }
+      double[] y = ds.GetDoubleValues(problemData.TargetVariable, rows).ToArray();
+
 
       // numeric constants in the tree become variables for constant opt
@@ -201 +215 @@
       // variable name, variable value (for factor vars) and lag as a DataForVariable object.
       // A dictionary is used to find parameters
-      double[] initialConstants;
-      var parameters = new List<TreeToAutoDiffTermConverter.DataForVariable>();
-
       TreeToAutoDiffTermConverter.ParametricFunction func;
       TreeToAutoDiffTermConverter.ParametricFunctionGradient func_grad;
-      if (!TreeToAutoDiffTermConverter.TryConvertToAutoDiff(tree, updateVariableWeights, applyLinearScaling, out parameters, out initialConstants, out func, out func_grad))
+
+      if (!TreeToAutoDiffTermConverter.TryConvertToAutoDiff(tree, updateVariableWeights, applyLinearScaling, parameters, out func, out func_grad))
         throw new NotSupportedException("Could not optimize constants of symbolic expression tree due to not supported symbols used in the tree.");
       if (parameters.Count == 0) return 0.0; // gkronber: constant expressions always have a R² of 0.0 
-      //var parameterEntries = parameters.ToArray(); // order of entries must be the same for x
+
+      var initialConstants = ConstantsOptimization.Util.ExtractConstants(tree);
 
       //extract inital constants
@@ -231 +244 @@
       int retVal;
 
-      IDataset ds = problemData.Dataset;
-      var x = ConstantsOptimization.Util.ExtractData(ds, parameters, rows);
-      double[] y = ds.GetDoubleValues(problemData.TargetVariable, rows).ToArray();
+
       int n = x.GetLength(0);
       int m = x.GetLength(1);
@@ -263 +274 @@
           var tmp = new double[c.Length - 2];
           Array.Copy(c, 2, tmp, 0, tmp.Length);
-          UpdateConstants(tree, tmp, updateVariableWeights);
-        } else UpdateConstants(tree, c, updateVariableWeights);
+          ConstantsOptimization.Util.UpdateConstants(tree, tmp);
+        } else ConstantsOptimization.Util.UpdateConstants(tree, c);
       }
       var quality = SymbolicRegressionSingleObjectivePearsonRSquaredEvaluator.Calculate(interpreter, tree, lowerEstimationLimit, upperEstimationLimit, problemData, rows, applyLinearScaling);
 
-      if (!updateConstantsInTree) UpdateConstants(tree, initialConstants, updateVariableWeights);
+      if (!updateConstantsInTree)
+        ConstantsOptimization.Util.UpdateConstants(tree, initialConstants);
 
       if (originalQuality - quality > 0.001 || double.IsNaN(quality)) {
-        UpdateConstants(tree, initialConstants, updateVariableWeights);
+        ConstantsOptimization.Util.UpdateConstants(tree, initialConstants);
         return originalQuality;
       }
       return quality;
-    }
-
-    private static void UpdateConstants(ISymbolicExpressionTree tree, double[] constants, bool updateVariableWeights) {
-      int i = 0;
-      foreach (var node in tree.Root.IterateNodesPrefix().OfType<SymbolicExpressionTreeTerminalNode>()) {
-        ConstantTreeNode constantTreeNode = node as ConstantTreeNode;
-        VariableTreeNodeBase variableTreeNodeBase = node as VariableTreeNodeBase;
-        FactorVariableTreeNode factorVarTreeNode = node as FactorVariableTreeNode;
-        if (constantTreeNode != null)
-          constantTreeNode.Value = constants[i++];
-        else if (updateVariableWeights && variableTreeNodeBase != null)
-          variableTreeNodeBase.Weight = constants[i++];
-        else if (factorVarTreeNode != null) {
-          for (int j = 0; j < factorVarTreeNode.Weights.Length; j++)
-            factorVarTreeNode.Weights[j] = constants[i++];
-        }
-      }
     }