Changeset 9360 for trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceFunctions

Timestamp:

04/15/13 15:07:16 (11 years ago)

Author:

gkronber

Message:

#1902: implemented neural network covariance function plus test case (comparison with GPML) for Gaussian processes

File:

• trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceFunctions/CovarianceNeuralNetwork.cs (modified) (7 diffs)

trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceFunctions/CovarianceNeuralNetwork.cs

@@ -40 +40 @@
     }
 
-    public IValueParameter<DoubleValue> InverseLengthParameter {
-      get { return (IValueParameter<DoubleValue>)Parameters["InverseLength"]; }
+    public IValueParameter<DoubleValue> LengthParameter {
+      get { return (IValueParameter<DoubleValue>)Parameters["Length"]; }
     }
 
@@ -59 +59 @@
 
       Parameters.Add(new OptionalValueParameter<DoubleValue>("Scale", "The scale parameter."));
-      Parameters.Add(new OptionalValueParameter<DoubleValue>("InverseLength", "The inverse length parameter."));
+      Parameters.Add(new OptionalValueParameter<DoubleValue>("Length", "The length parameter."));
     }
 
@@ -69 +69 @@
       return
         (ScaleParameter.Value != null ? 0 : 1) +
-        (InverseLengthParameter.Value != null ? 0 : 1);
+        (LengthParameter.Value != null ? 0 : 1);
     }
 
     public void SetParameter(double[] p) {
-      double scale, inverseLength;
-      GetParameterValues(p, out scale, out inverseLength);
+      double scale, length;
+      GetParameterValues(p, out scale, out length);
       ScaleParameter.Value = new DoubleValue(scale);
-      InverseLengthParameter.Value = new DoubleValue(inverseLength);
+      LengthParameter.Value = new DoubleValue(length);
     }
 
 
-    private void GetParameterValues(double[] p, out double scale, out double inverseLength) {
+    private void GetParameterValues(double[] p, out double scale, out double length) {
       // gather parameter values
       int c = 0;
-      if (InverseLengthParameter.Value != null) {
-        inverseLength = InverseLengthParameter.Value.Value;
+      if (LengthParameter.Value != null) {
+        length = LengthParameter.Value.Value;
       } else {
-        inverseLength = 1.0 / Math.Exp(p[c]);
+        length = Math.Exp(2 * p[c]);
         c++;
       }
@@ -105 +105 @@
     private static Func<Term, UnaryFunc> sqrt = UnaryFunc.Factory(
       x => Math.Sqrt(x),
-      x => 1 / 2 * Math.Sqrt(x));
+      x => 1 / (2 * Math.Sqrt(x)));
 
     public ParameterizedCovarianceFunction GetParameterizedCovarianceFunction(double[] p, IEnumerable<int> columnIndices) {
-      double inverseLength, scale;
-      GetParameterValues(p, out scale, out inverseLength);
+      double length, scale;
+      GetParameterValues(p, out scale, out length);
       // create functions
       AutoDiff.Variable p0 = new AutoDiff.Variable();
       AutoDiff.Variable p1 = new AutoDiff.Variable();
-      var invL = 1.0 / TermBuilder.Exp(p0);
-      var s = TermBuilder.Exp(2 * p1);
+      var l = TermBuilder.Exp(2.0 * p0);
+      var s = TermBuilder.Exp(2.0 * p1);
       AutoDiff.Variable[] x1 = new AutoDiff.Variable[columnIndices.Count()];
       AutoDiff.Variable[] x2 = new AutoDiff.Variable[columnIndices.Count()];
-      AutoDiff.Term sx = invL;
-      AutoDiff.Term s1 = invL;
-      AutoDiff.Term s2 = invL;
+      AutoDiff.Term sx = 1;
+      AutoDiff.Term s1 = 1;
+      AutoDiff.Term s2 = 1;
       foreach (var k in columnIndices) {
         x1[k] = new AutoDiff.Variable();
         x2[k] = new AutoDiff.Variable();
-        sx += x1[k] * invL * x2[k];
-        s1 += x1[k] * invL * x1[k];
-        s2 += x2[k] * invL * x2[k];
+        sx += x1[k] * x2[k];
+        s1 += x1[k] * x1[k];
+        s2 += x2[k] * x2[k];
       }
 
       var parameter = x1.Concat(x2).Concat(new AutoDiff.Variable[] { p0, p1 }).ToArray();
       var values = new double[x1.Length + x2.Length + 2];
-      var c = (s * asin(2 * sx / (sqrt((1 + 2 * s1) * (1 + 2 * s2))))).Compile(parameter);
+      var c = (s * asin(sx / (sqrt((l + s1) * (l + s2))))).Compile(parameter);
 
       var cov = new ParameterizedCovarianceFunction();
@@ -143 +143 @@
           k++;
         }
-        values[k] = Math.Log(1.0 / inverseLength);
-        values[k + 1] = Math.Log(scale) / 2.0;
+        values[k] = Math.Log(Math.Sqrt(length));
+        values[k + 1] = Math.Log(Math.Sqrt(scale));
         return c.Evaluate(values);
       };
@@ -157 +157 @@
           k++;
         }
-        values[k] = Math.Log(1.0 / inverseLength);
-        values[k + 1] = Math.Log(scale) / 2.0;
+        values[k] = Math.Log(Math.Sqrt(length));
+        values[k + 1] = Math.Log(Math.Sqrt(scale));
         return c.Evaluate(values);
       };
@@ -171 +171 @@
           k++;
         }
-        values[k] = Math.Log(1.0 / inverseLength);
-        values[k + 1] = Math.Log(scale) / 2.0;
+        values[k] = Math.Log(Math.Sqrt(length));
+        values[k + 1] = Math.Log(Math.Sqrt(scale));
         return c.Differentiate(values).Item1.Skip(columnIndices.Count() * 2);
       };