source: trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceFunctions/CovarianceSquaredExponentialArd.cs @ 13784

Last change on this file since 13784 was 13784, checked in by pfleck, 5 years ago

#2591 Made the creation of a GaussianProcessModel faster by avoiding additional iterators during calculation of the hyperparameter gradients.
The gradients of the hyperparameters are now calculated in one sweep and returned as IList, instead of returning an iterator (with yield return).
This avoids a large amount of Move-calls of the iterator, especially for covariance functions with a lot of hyperparameters.
Besides, the signature of the CovarianceGradientFunctionDelegate is changed, to return an IList instead of an IEnumerable to avoid unnececary ToList or ToArray calls.

File size: 5.9 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2015 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
4 *
5 * This file is part of HeuristicLab.
6 *
7 * HeuristicLab is free software: you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation, either version 3 of the License, or
10 * (at your option) any later version.
11 *
12 * HeuristicLab is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
19 */
20#endregion
21
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Core;
27using HeuristicLab.Data;
28using HeuristicLab.Parameters;
29using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
30
31namespace HeuristicLab.Algorithms.DataAnalysis {
32  [StorableClass]
33  [Item(Name = "CovarianceSquaredExponentialArd", Description = "Squared exponential covariance function with automatic relevance determination for Gaussian processes.")]
34  public sealed class CovarianceSquaredExponentialArd : ParameterizedNamedItem, ICovarianceFunction {
35    public IValueParameter<DoubleValue> ScaleParameter {
36      get { return (IValueParameter<DoubleValue>)Parameters["Scale"]; }
37    }
38
39    public IValueParameter<DoubleArray> InverseLengthParameter {
40      get { return (IValueParameter<DoubleArray>)Parameters["InverseLength"]; }
41    }
42    private bool HasFixedInverseLengthParameter {
43      get { return InverseLengthParameter.Value != null; }
44    }
45    private bool HasFixedScaleParameter {
46      get { return ScaleParameter.Value != null; }
47    }
48
49    [StorableConstructor]
50    private CovarianceSquaredExponentialArd(bool deserializing) : base(deserializing) { }
51    private CovarianceSquaredExponentialArd(CovarianceSquaredExponentialArd original, Cloner cloner)
52      : base(original, cloner) {
53    }
54    public CovarianceSquaredExponentialArd()
55      : base() {
56      Name = ItemName;
57      Description = ItemDescription;
58
59      Parameters.Add(new OptionalValueParameter<DoubleValue>("Scale", "The scale parameter of the squared exponential covariance function with ARD."));
60      Parameters.Add(new OptionalValueParameter<DoubleArray>("InverseLength", "The inverse length parameter for automatic relevance determination."));
61    }
62
63    public override IDeepCloneable Clone(Cloner cloner) {
64      return new CovarianceSquaredExponentialArd(this, cloner);
65    }
66
67    public int GetNumberOfParameters(int numberOfVariables) {
68      return
69        (HasFixedScaleParameter ? 0 : 1) +
70        (HasFixedInverseLengthParameter ? 0 : numberOfVariables);
71    }
72
73    public void SetParameter(double[] p) {
74      double scale;
75      double[] inverseLength;
76      GetParameterValues(p, out scale, out inverseLength);
77      ScaleParameter.Value = new DoubleValue(scale);
78      InverseLengthParameter.Value = new DoubleArray(inverseLength);
79    }
80
81    private void GetParameterValues(double[] p, out double scale, out double[] inverseLength) {
82      int c = 0;
83      // gather parameter values
84      if (HasFixedInverseLengthParameter) {
85        inverseLength = InverseLengthParameter.Value.ToArray();
86      } else {
87        int length = p.Length;
88        if (!HasFixedScaleParameter) length--;
89        inverseLength = p.Select(e => 1.0 / Math.Exp(e)).Take(length).ToArray();
90        c += inverseLength.Length;
91      }
92      if (HasFixedScaleParameter) {
93        scale = ScaleParameter.Value.Value;
94      } else {
95        scale = Math.Exp(2 * p[c]);
96        c++;
97      }
98      if (p.Length != c) throw new ArgumentException("The length of the parameter vector does not match the number of free parameters for CovarianceSquaredExponentialArd", "p");
99    }
100
101    public ParameterizedCovarianceFunction GetParameterizedCovarianceFunction(double[] p, int[] columnIndices) {
102      double scale;
103      double[] inverseLength;
104      GetParameterValues(p, out scale, out inverseLength);
105      var fixedInverseLength = HasFixedInverseLengthParameter;
106      var fixedScale = HasFixedScaleParameter;
107      // create functions
108      var cov = new ParameterizedCovarianceFunction();
109      cov.Covariance = (x, i, j) => {
110        double d = i == j
111                 ? 0.0
112                 : Util.SqrDist(x, i, j, inverseLength, columnIndices);
113        return scale * Math.Exp(-d / 2.0);
114      };
115      cov.CrossCovariance = (x, xt, i, j) => {
116        double d = Util.SqrDist(x, i, xt, j, inverseLength, columnIndices);
117        return scale * Math.Exp(-d / 2.0);
118      };
119      cov.CovarianceGradient = (x, i, j) => GetGradient(x, i, j, columnIndices, scale, inverseLength, fixedInverseLength, fixedScale);
120      return cov;
121    }
122
123    // order of returned gradients must match the order in GetParameterValues!
124    private static IList<double> GetGradient(double[,] x, int i, int j, int[] columnIndices, double scale, double[] inverseLength,
125      bool fixedInverseLength, bool fixedScale) {
126      double d = i == j
127                   ? 0.0
128                   : Util.SqrDist(x, i, j, inverseLength, columnIndices);
129
130      int k = 0;
131      var g = new List<double>((!fixedInverseLength ? columnIndices.Length : 0) + (!fixedScale ? 1 : 0));
132      if (!fixedInverseLength) {
133        for (int c = 0; c < columnIndices.Length; c++) {
134          var columnIndex = columnIndices[c];
135          double sqrDist = Util.SqrDist(x[i, columnIndex] * inverseLength[k], x[j, columnIndex] * inverseLength[k]);
136          g.Add(scale * Math.Exp(-d / 2.0) * sqrDist);
137          k++;
138        }
139      }
140      if (!fixedScale) g.Add(2.0 * scale * Math.Exp(-d / 2.0));
141      return g;
142    }
143  }
144}
Note: See TracBrowser for help on using the repository browser.