source: trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/GaussianProcess/CovarianceFunctions/CovariancePeriodic.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: 6.5 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 HeuristicLab.Common;
25using HeuristicLab.Core;
26using HeuristicLab.Data;
27using HeuristicLab.Parameters;
28using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
29
30namespace HeuristicLab.Algorithms.DataAnalysis {
31  [StorableClass]
32  [Item(Name = "CovariancePeriodic", Description = "Periodic covariance function for Gaussian processes.")]
33  public sealed class CovariancePeriodic : ParameterizedNamedItem, ICovarianceFunction {
34
35    public IValueParameter<DoubleValue> ScaleParameter {
36      get { return (IValueParameter<DoubleValue>)Parameters["Scale"]; }
37    }
38
39    public IValueParameter<DoubleValue> InverseLengthParameter {
40      get { return (IValueParameter<DoubleValue>)Parameters["InverseLength"]; }
41    }
42
43    public IValueParameter<DoubleValue> PeriodParameter {
44      get { return (IValueParameter<DoubleValue>)Parameters["Period"]; }
45    }
46
47    private bool HasFixedScaleParameter {
48      get { return ScaleParameter.Value != null; }
49    }
50    private bool HasFixedInverseLengthParameter {
51      get { return InverseLengthParameter.Value != null; }
52    }
53    private bool HasFixedPeriodParameter {
54      get { return PeriodParameter.Value != null; }
55    }
56
57
58    [StorableConstructor]
59    private CovariancePeriodic(bool deserializing) : base(deserializing) { }
60    private CovariancePeriodic(CovariancePeriodic original, Cloner cloner)
61      : base(original, cloner) {
62    }
63
64    public CovariancePeriodic()
65      : base() {
66      Name = ItemName;
67      Description = ItemDescription;
68
69      Parameters.Add(new OptionalValueParameter<DoubleValue>("Scale", "The scale of the periodic covariance function."));
70      Parameters.Add(new OptionalValueParameter<DoubleValue>("InverseLength", "The inverse length parameter for the periodic covariance function."));
71      Parameters.Add(new OptionalValueParameter<DoubleValue>("Period", "The period parameter for the periodic covariance function."));
72    }
73
74    public override IDeepCloneable Clone(Cloner cloner) {
75      return new CovariancePeriodic(this, cloner);
76    }
77
78    public int GetNumberOfParameters(int numberOfVariables) {
79      return (HasFixedScaleParameter ? 0 : 1) +
80       (HasFixedPeriodParameter ? 0 : 1) +
81       (HasFixedInverseLengthParameter ? 0 : 1);
82    }
83
84    public void SetParameter(double[] p) {
85      double scale, inverseLength, period;
86      GetParameterValues(p, out scale, out period, out inverseLength);
87      ScaleParameter.Value = new DoubleValue(scale);
88      PeriodParameter.Value = new DoubleValue(period);
89      InverseLengthParameter.Value = new DoubleValue(inverseLength);
90    }
91
92
93    private void GetParameterValues(double[]
94      p, out double scale, out double period, out double inverseLength) {
95      // gather parameter values
96      int c = 0;
97      if (HasFixedInverseLengthParameter) {
98        inverseLength = InverseLengthParameter.Value.Value;
99      } else {
100        inverseLength = 1.0 / Math.Exp(p[c]);
101        c++;
102      }
103      if (HasFixedPeriodParameter) {
104        period = PeriodParameter.Value.Value;
105      } else {
106        period = Math.Exp(p[c]);
107        c++;
108      }
109      if (HasFixedScaleParameter) {
110        scale = ScaleParameter.Value.Value;
111      } else {
112        scale = Math.Exp(2 * p[c]);
113        c++;
114      }
115      if (p.Length != c) throw new ArgumentException("The length of the parameter vector does not match the number of free parameters for CovariancePeriodic", "p");
116    }
117
118    public ParameterizedCovarianceFunction GetParameterizedCovarianceFunction(double[] p, int[] columnIndices) {
119      double inverseLength, period, scale;
120      GetParameterValues(p, out scale, out period, out inverseLength);
121      var fixedInverseLength = HasFixedInverseLengthParameter;
122      var fixedPeriod = HasFixedPeriodParameter;
123      var fixedScale = HasFixedScaleParameter;
124      // create functions
125      var cov = new ParameterizedCovarianceFunction();
126      cov.Covariance = (x, i, j) => {
127        double k = i == j ? 0.0 : GetDistance(x, x, i, j, columnIndices);
128        k = Math.PI * k / period;
129        k = Math.Sin(k) * inverseLength;
130        k = k * k;
131
132        return scale * Math.Exp(-2.0 * k);
133      };
134      cov.CrossCovariance = (x, xt, i, j) => {
135        double k = GetDistance(x, xt, i, j, columnIndices);
136        k = Math.PI * k / period;
137        k = Math.Sin(k) * inverseLength;
138        k = k * k;
139
140        return scale * Math.Exp(-2.0 * k);
141      };
142      cov.CovarianceGradient = (x, i, j) => GetGradient(x, i, j, columnIndices, scale, period, inverseLength, fixedInverseLength, fixedPeriod, fixedScale);
143      return cov;
144    }
145
146    private static IList<double> GetGradient(double[,] x, int i, int j, int[] columnIndices, double scale, double period, double inverseLength,
147      bool fixedInverseLength, bool fixedPeriod, bool fixedScale) {
148      double k = i == j ? 0.0 : Math.PI * GetDistance(x, x, i, j, columnIndices) / period;
149      double gradient = Math.Sin(k) * inverseLength;
150      gradient *= gradient;
151      var g = new List<double>(3);
152      if (!fixedInverseLength)
153        g.Add(4.0 * scale * Math.Exp(-2.0 * gradient) * gradient);
154      if (!fixedPeriod) {
155        double r = Math.Sin(k) * inverseLength;
156        g.Add(2.0 * k * scale * Math.Exp(-2 * r * r) * Math.Sin(2 * k) * inverseLength * inverseLength);
157      }
158      if (!fixedScale)
159        g.Add(2.0 * scale * Math.Exp(-2 * gradient));
160      return g;
161    }
162
163    private static double GetDistance(double[,] x, double[,] xt, int i, int j, int[] columnIndices) {
164      return Math.Sqrt(Util.SqrDist(x, i, xt, j, columnIndices, 1));
165    }
166  }
167}
Note: See TracBrowser for help on using the repository browser.