Changeset 15457 for branches/MathNetNumerics-Exploration-2789/HeuristicLab.Algorithms.DataAnalysis.Experimental/Splines.cs

Timestamp:

11/07/17 17:03:02 (6 years ago)

Author:

gkronber

Message:

#2789 added Finbarr O'Sullivan smoothing spline code

File:

• branches/MathNetNumerics-Exploration-2789/HeuristicLab.Algorithms.DataAnalysis.Experimental/Splines.cs (modified) (10 diffs)

branches/MathNetNumerics-Exploration-2789/HeuristicLab.Algorithms.DataAnalysis.Experimental/Splines.cs

@@ -70 +70 @@
           "B-Spline Smoothing",
           "Penalized Regression Spline (alglib)",
-          "CUBGCV"
+          "CUBGCV",
+          "Finnbar O'Sullivan (sbart)"
       }.Select(s => new StringValue(s)));
 
@@ -172 +173 @@
               break;
             }
+          case "Finnbar O'Sullivan (sbart)": {
+              SBART.SBART_Report report;
+              var lambda = ((IValueParameter<DoubleValue>)Parameters["Lambda"]).Value.Value; // controls extent of smoothing
+              var model =
+                SBART.CalculateSBART(x, y,
+                Problem.ProblemData.TargetVariable, inputVars, (float)lambda, out report);
+              var targetVar = Problem.ProblemData.TargetVariable;
+              var problemData = (IRegressionProblemData)Problem.ProblemData.Clone();
+              Results.Add(new Result("Solution", model.CreateRegressionSolution(problemData)));
+              Results.Add(new Result("GCV", new DoubleValue(report.gcv)));
+              Results.Add(new Result("Smoothing parameter (optimized for GCV)", new DoubleValue(report.smoothingParameter)));
+              break;
+
+            }
           case "B-Spline Smoothing": {
               BSplineSmoothing(x, y, inputVars);
@@ -205 +220 @@
     public static IRegressionEnsembleModel CalculatePenalizedRegressionSpline(double[] x, double[] y, double lambda, string targetVar, string[] inputVars,
       out double avgTrainRMSE,
-      out double cvRMSE) {
+      out double cvRMSE, out double[] predictions) {
       int info;
       alglib.spline1dinterpolant interpolant;
@@ -218 +233 @@
       double[] x_loo = new double[(folds - 1) * foldSize];
       double[] y_loo = new double[(folds - 1) * foldSize];
+      double[] w_loo = Enumerable.Repeat(1.0, x_loo.Length).ToArray();
 
       var models = new List<IRegressionModel>();
@@ -225 +241 @@
       int nTest = 0;
 
+      predictions = new double[y.Length];
 
       for (int i = 0; i < folds; i++) {
@@ -232 +249 @@
           Array.Copy(y, (i + 1) * foldSize, y_loo, i * foldSize, (folds - i - 1) * foldSize);
         }
-        alglib.spline1dfitpenalized(x_loo, y_loo, n, lambda, out info, out interpolant, out rep);
-        Debug.Assert(y_loo.All(yi => yi != 0.0));
-        Debug.Assert(x_loo.All(xi => xi != 0.0));
+        alglib.spline1dfitpenalizedw(x_loo, y_loo, w_loo, n, lambda, out info, out interpolant, out rep);
+        //Debug.Assert(y_loo.All(yi => yi != 0.0));
+        //Debug.Assert(x_loo.All(xi => xi != 0.0));
 
         // training error
         for (int j = 0; j < x_loo.Length; j++) {
-          var y_pred = alglib.spline1dcalc(interpolant, x[i]);
+          var y_pred = alglib.spline1dcalc(interpolant, x[j]);
           double res = y[j] - y_pred;
           sumTrainSE += res * res;
@@ -245 +262 @@
         // test
         for (int j = i * foldSize; j < (i + 1) * foldSize; j++) {
-          var y_pred = alglib.spline1dcalc(interpolant, x[i]);
-          double res = y[j] - y_pred;
+          predictions[j] = alglib.spline1dcalc(interpolant, x[j]);
+          double res = y[j] - predictions[j];
           sumTestSE += res * res;
           nTest++;
@@ -630 +647 @@
       int n = x.Length;
       var range = x[n - 1] - x[0];
-      var binSize = range / n / 20;
+      var binSize = range / n / 10;
       {
         // binning
@@ -759 +776 @@
 
     public double GetEstimatedValue(double xx) {
+      xx = (xx - offset) * scale;
       int n = a.Length;
       if (xx <= x[1]) {
@@ -791 +809 @@
         product = product.Zip(dataset.GetDoubleValues(factor, rows), (pi, fi) => pi * fi).ToArray();
       }
-      foreach (var xx in product.Select(xi => (xi - offset) * scale)) {
+      foreach (var xx in product) {
         yield return GetEstimatedValue(xx);
       }