Changeset 17740 for branches/3022-FastFunctionExtraction/FFX/BFUtils.cs

Timestamp:

08/31/20 12:06:20 (4 years ago)

Author:

lleko

Message:

#3022 add normalization

File:

• branches/3022-FastFunctionExtraction/FFX/BFUtils.cs (modified) (4 diffs)

branches/3022-FastFunctionExtraction/FFX/BFUtils.cs

@@ -1 +1 @@
 using HeuristicLab.Algorithms.DataAnalysis.Glmnet;
+using HeuristicLab.Common;
 using HeuristicLab.Data;
 using HeuristicLab.Problems.DataAnalysis;
@@ -18 +19 @@
             if (approach.AllowHinge) {
                 // only allow hinge functions for features with exponent = 1 (deemed too complex otherwise)
-                var linearSimpleBasisFuncs = simpleBasisFuncs.Where(simpleBf => simpleBf.Exponent == 1);
+                var linearSimpleBasisFuncs = simpleBasisFuncs.Where(simpleBf => simpleBf.Exponent == 1 && simpleBf.Operator.Equals(NonlinearOperator.None));
                 simpleBasisFuncs = simpleBasisFuncs.Concat(CreateHingeBases(data, linearSimpleBasisFuncs, approach.MinHingeThr, approach.MaxHingeThr, approach.NumHingeThrs));
             }
@@ -88 +89 @@
         // the importance of a basis function is measured by the absolute value of its coefficient when optimized on the data
         public static IEnumerable<ISimpleBasisFunction> OrderBasisFuncsByImportance(IRegressionProblemData data, IList<ISimpleBasisFunction> candidateFunctions) {
-            var elnetData = PrepareData(Normalize(data), candidateFunctions);
+            var elnetData = PrepareData(Normalize(data, out _, out _, out _, out _), candidateFunctions);
             var coeff = ElasticNetLinearRegression.CalculateModelCoefficients(elnetData, 0, 0, out var trainNMSE, out var testNMSE); // LS-fit
             var intercept = coeff.Last();
@@ -177 +178 @@
         }
 
-        private static IRegressionProblemData Normalize(IRegressionProblemData data)
-            => new RegressionProblemData(Normalize(data.Dataset), data.AllowedInputVariables, data.TargetVariable);
+        public static IRegressionProblemData Normalize(IRegressionProblemData data, out double[] X_avgs, out double[] X_stds, out double y_avg, out double y_std) {
+            X_avgs = data.AllowedInputVariables
+                .Select(varname => data.Dataset.GetDoubleValues(varname)
+                    .Average())
+                .ToArray();
+            X_stds = data.AllowedInputVariables
+                .Select(varname => data.Dataset.GetDoubleValues(varname)
+                    .StandardDeviationPop())
+                .ToArray();
+            for (int i = 0; i < X_stds.Length; i++) {
+                if (X_stds[i] == 0) X_stds[i] = 1;
+            }
+            y_avg = data.TargetVariableValues.Average();
+            y_std = data.TargetVariableValues.StandardDeviationPop();
+            if (y_std == 0) y_std = 1;
+            var temp = Normalize(data.Dataset);
+            var ans = new RegressionProblemData(Normalize(data.Dataset), data.AllowedInputVariables, data.TargetVariable);
+            return ans;
+        }
 
         // return a normalized version of IDataset ds