source: branches/HeuristicLab.TimeSeries/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceSEard.cs @ 8477

Last change on this file since 8477 was 8477, checked in by mkommend, 9 years ago

#1081:

  • Added autoregressive target variable Symbol
  • Merged trunk changes into the branch.
File size: 4.8 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2012 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.Linq;
24using HeuristicLab.Common;
25using HeuristicLab.Core;
26using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
27
28namespace HeuristicLab.Algorithms.DataAnalysis {
29  [StorableClass]
30  [Item(Name = "CovarianceSEard", Description = "Squared exponential covariance function with automatic relevance determination for Gaussian processes.")]
31  public class CovarianceSEard : Item, ICovarianceFunction {
32    [Storable]
33    private double[,] x;
34    [Storable]
35    private double[,] xt;
36    [Storable]
37    private double sf2;
38    public double Scale { get { return sf2; } }
39
40    [Storable]
41    private double[] l;
42    public double[] Length {
43      get {
44        if (l == null) return new double[0];
45        var copy = new double[l.Length];
46        Array.Copy(l, copy, copy.Length);
47        return copy;
48      }
49    }
50
51    private double[,] sd;
52    private bool symmetric;
53
54    public int GetNumberOfParameters(int numberOfVariables) {
55      return numberOfVariables + 1;
56    }
57    [StorableConstructor]
58    protected CovarianceSEard(bool deserializing) : base(deserializing) { }
59    protected CovarianceSEard(CovarianceSEard original, Cloner cloner)
60      : base(original, cloner) {
61      if (original.x != null) {
62        this.x = new double[original.x.GetLength(0), original.x.GetLength(1)];
63        Array.Copy(original.x, this.x, x.Length);
64
65        this.xt = new double[original.xt.GetLength(0), original.xt.GetLength(1)];
66        Array.Copy(original.xt, this.xt, xt.Length);
67
68        this.sd = new double[original.sd.GetLength(0), original.sd.GetLength(1)];
69        Array.Copy(original.sd, this.sd, sd.Length);
70
71        this.l = new double[original.l.Length];
72        Array.Copy(original.l, this.l, l.Length);
73      }
74      this.sf2 = original.sf2;
75      this.symmetric = original.symmetric;
76    }
77    public CovarianceSEard()
78      : base() {
79    }
80
81    public override IDeepCloneable Clone(Cloner cloner) {
82      return new CovarianceSEard(this, cloner);
83    }
84
85    public void SetParameter(double[] hyp) {
86      this.l = hyp.Take(hyp.Length - 1).Select(Math.Exp).ToArray();
87      this.sf2 = Math.Exp(2 * hyp[hyp.Length - 1]);
88      // sf2 = Math.Min(10E6, sf2); // upper limit for the scale
89
90      sd = null;
91    }
92
93    public void SetData(double[,] x) {
94      SetData(x, x);
95      this.symmetric = true;
96    }
97
98    public void SetData(double[,] x, double[,] xt) {
99      this.x = x;
100      this.xt = xt;
101      this.symmetric = false;
102
103      sd = null;
104    }
105
106    public double GetCovariance(int i, int j) {
107      if (sd == null) CalculateSquaredDistances();
108      return sf2 * Math.Exp(-sd[i, j] / 2.0);
109    }
110
111    public double GetGradient(int i, int j, int k) {
112      if (k < l.Length) {
113        double sqrDist = Util.SqrDist(x[i, k] / l[k], xt[j, k] / l[k]);
114        return sf2 * Math.Exp(-sd[i, j] / 2.0) * sqrDist;
115      } else if (k == l.Length) {
116        return 2.0 * sf2 * Math.Exp(-sd[i, j] / 2.0);
117      } else {
118        throw new ArgumentException("CovarianceSEard has dimension+1 hyperparameters.", "k");
119      }
120    }
121
122
123    private void CalculateSquaredDistances() {
124      if (x.GetLength(1) != xt.GetLength(1)) throw new InvalidOperationException();
125      int rows = x.GetLength(0);
126      int cols = xt.GetLength(0);
127      sd = new double[rows, cols];
128      if (symmetric) {
129        for (int i = 0; i < rows; i++) {
130          for (int j = i; j < cols; j++) {
131            sd[i, j] = Util.SqrDist(Util.GetRow(x, i).Select((e, k) => e / l[k]),
132                                    Util.GetRow(xt, j).Select((e, k) => e / l[k]));
133            sd[j, i] = sd[i, j];
134          }
135        }
136      } else {
137        for (int i = 0; i < rows; i++) {
138          for (int j = 0; j < cols; j++) {
139            sd[i, j] = Util.SqrDist(Util.GetRow(x, i).Select((e, k) => e / l[k]),
140                                    Util.GetRow(xt, j).Select((e, k) => e / l[k]));
141          }
142        }
143      }
144    }
145  }
146}
Note: See TracBrowser for help on using the repository browser.