Changeset 10422 for branches/HeuristicLab.Problems.GaussianProcessTuning/GaussianProcessDemo/Form1.cs

Timestamp:

01/29/14 14:34:44 (10 years ago)

Author:

gkronber

Message:

#1967 added Cholesky decomposition for Toeplitz matrices to allow sampling from one-dimensional Gaussian processes

File:

• branches/HeuristicLab.Problems.GaussianProcessTuning/GaussianProcessDemo/Form1.cs (modified) (5 diffs)

branches/HeuristicLab.Problems.GaussianProcessTuning/GaussianProcessDemo/Form1.cs

@@ -4 +4 @@
 using System.Data;
 using System.Drawing;
+using System.Globalization;
 using System.Linq;
 using System.Text;
@@ -28 +29 @@
 
       var sum = new CovarianceSum();
-      var t = new CovarianceNeuralNetwork();
+      var t = new CovarianceSquaredExponentialIso();
       sum.Terms.Add(t);
-      sum.Terms.Add(new CovarianceNoise());
+      //sum.Terms.Add(new CovarianceNoise());
       this.covFunction = sum;
       UpdateSliders();
@@ -52 +53 @@
 
     private void InitData() {
-      int n = 200;
+      int n = 5000;
       data = new List<List<double>>();
-      data.Add(ValueGenerator.GenerateSteps(0, 1, 1.0 / n).ToList());
+      data.Add(Enumerable.Range(0, n).Select(e => e / 200.0).ToList());
 
       // sample from GP
       var normalRand = new NormalDistributedRandom(random, 0, 1);
-      alpha = (from i in Enumerable.Range(0, n + 1)
+      alpha = (from i in Enumerable.Range(0, n)
                select normalRand.NextDouble()).ToArray();
     }
@@ -72 +73 @@
       }
 
-      var trainingData = new List<List<double>>();
-      var trainingIndices = RandomEnumerable.SampleRandomWithoutRepetition(Enumerable.Range(0, y.Count), random, 10);
-      var trainingY = trainingIndices.Select(i => y[i]).ToList();
-      var trainingX = trainingIndices.Select(i => data[0][i]).ToList();
-      trainingData.Add(trainingY);
-      trainingData.Add(trainingX);
-      
-      //chart1.Series[2].Points.Clear();
-      //foreach (var p in trainingY.Zip(trainingX, (t, x) => new { t, x })) {
-      //  chart1.Series[2].Points.AddXY(p.x, p.t);
+      //var trainingData = new List<List<double>>();
+      //var trainingIndices = RandomEnumerable.SampleRandomWithoutRepetition(Enumerable.Range(0, y.Count), random, 10);
+      //var trainingY = trainingIndices.Select(i => y[i]).ToList();
+      //var trainingX = trainingIndices.Select(i => data[0][i]).ToList();
+      //trainingData.Add(trainingY);
+      //trainingData.Add(trainingX);
+      //
+      ////chart1.Series[2].Points.Clear();
+      ////foreach (var p in trainingY.Zip(trainingX, (t, x) => new { t, x })) {
+      ////  chart1.Series[2].Points.AddXY(p.x, p.t);
+      ////}
+      //
+      //var allData = new List<List<double>>();
+      //allData.Add(y);
+      //allData.Add(data[0]);
+      //var variableNames = new string[] { "y", "x" };
+      //var fullDataSet = new Dataset(variableNames, allData);
+      //var trainingDataSet = new Dataset(variableNames, trainingData);
+      //var trainingRows = Enumerable.Range(0, trainingIndices.Count());
+      //var fullRows = Enumerable.Range(0, data[0].Count);
+      //var correctModel = new GaussianProcessModel(fullDataSet, variableNames.First(), variableNames.Skip(1), fullRows, hyp, new MeanZero(),
+      //                                          (ICovarianceFunction)covFunction.Clone());
+      //var yPred = correctModel.GetEstimatedValues(fullDataSet, fullRows);
+      //chart1.Series[1].Points.Clear();
+      //foreach (var p in yPred.Zip(data[0], (t, x) => new { t, x })) {
+      //  chart1.Series[1].Points.AddXY(p.x, p.t);
       //}
-
-      var allData = new List<List<double>>();
-      allData.Add(y);
-      allData.Add(data[0]);
-      var variableNames = new string[] { "y", "x" };
-      var fullDataSet = new Dataset(variableNames, allData);
-      var trainingDataSet = new Dataset(variableNames, trainingData);
-      var trainingRows = Enumerable.Range(0, trainingIndices.Count());
-      var fullRows = Enumerable.Range(0, data[0].Count);
-      var correctModel = new GaussianProcessModel(fullDataSet, variableNames.First(), variableNames.Skip(1), fullRows, hyp, new MeanZero(),
-                                                (ICovarianceFunction)covFunction.Clone());
-      var yPred = correctModel.GetEstimatedValues(fullDataSet, fullRows);
-      chart1.Series[1].Points.Clear();
-      foreach (var p in yPred.Zip(data[0], (t, x) => new { t, x })) {
-        chart1.Series[1].Points.AddXY(p.x, p.t);
-      }
     }
 
@@ -124 +125 @@
     private string DataToText() {
       var str = new StringBuilder();
-      foreach (var p in chart1.Series[1].Points) {
-        str.AppendLine(p.XValue + "\t" + p.YValues.First());
+      foreach (var p in chart1.Series[0].Points) {
+        str.AppendLine(p.XValue.ToString(CultureInfo.InvariantCulture) + "\t" + p.YValues.First().ToString(CultureInfo.InvariantCulture));
       }
       return str.ToString();