[8565] | 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 |
|
---|
| 22 | using System;
|
---|
| 23 | using System.Collections.Generic;
|
---|
| 24 | using System.Linq;
|
---|
| 25 | using HeuristicLab.Common;
|
---|
| 26 | using HeuristicLab.Core;
|
---|
[8612] | 27 | using HeuristicLab.Data;
|
---|
[8565] | 28 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 29 |
|
---|
| 30 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
| 31 | [StorableClass]
|
---|
| 32 | [Item(Name = "CovarianceRQArd",
|
---|
| 33 | Description = "Rational quadratic covariance function with automatic relevance determination for Gaussian processes.")]
|
---|
[8612] | 34 | public sealed class CovarianceRQArd : ParameterizedNamedItem, ICovarianceFunction {
|
---|
[8565] | 35 | [Storable]
|
---|
| 36 | private double sf2;
|
---|
| 37 | [Storable]
|
---|
[8612] | 38 | private readonly HyperParameter<DoubleValue> scaleParameter;
|
---|
| 39 | public IValueParameter<DoubleValue> ScaleParameter {
|
---|
| 40 | get { return scaleParameter; }
|
---|
| 41 | }
|
---|
| 42 |
|
---|
| 43 | [Storable]
|
---|
[8565] | 44 | private double[] inverseLength;
|
---|
[8612] | 45 | [Storable]
|
---|
| 46 | private readonly HyperParameter<DoubleArray> inverseLengthParameter;
|
---|
| 47 | public IValueParameter<DoubleArray> InverseLengthParameter {
|
---|
| 48 | get { return inverseLengthParameter; }
|
---|
[8565] | 49 | }
|
---|
[8612] | 50 |
|
---|
[8565] | 51 | [Storable]
|
---|
[8612] | 52 | private double shape;
|
---|
| 53 | [Storable]
|
---|
| 54 | private readonly HyperParameter<DoubleValue> shapeParameter;
|
---|
| 55 | public IValueParameter<DoubleValue> ShapeParameter {
|
---|
| 56 | get { return shapeParameter; }
|
---|
| 57 | }
|
---|
[8565] | 58 |
|
---|
| 59 | [StorableConstructor]
|
---|
[8612] | 60 | private CovarianceRQArd(bool deserializing)
|
---|
[8565] | 61 | : base(deserializing) {
|
---|
| 62 | }
|
---|
| 63 |
|
---|
[8612] | 64 | private CovarianceRQArd(CovarianceRQArd original, Cloner cloner)
|
---|
[8565] | 65 | : base(original, cloner) {
|
---|
[8612] | 66 | this.scaleParameter = cloner.Clone(original.scaleParameter);
|
---|
[8565] | 67 | this.sf2 = original.sf2;
|
---|
[8612] | 68 |
|
---|
| 69 | this.inverseLengthParameter = cloner.Clone(original.inverseLengthParameter);
|
---|
| 70 | if (original.inverseLength != null) {
|
---|
| 71 | this.inverseLength = new double[original.inverseLength.Length];
|
---|
| 72 | Array.Copy(original.inverseLength, inverseLength, inverseLength.Length);
|
---|
| 73 | }
|
---|
| 74 |
|
---|
| 75 | this.shapeParameter = cloner.Clone(original.shapeParameter);
|
---|
| 76 | this.shape = original.shape;
|
---|
| 77 |
|
---|
| 78 | RegisterEvents();
|
---|
[8565] | 79 | }
|
---|
| 80 |
|
---|
| 81 | public CovarianceRQArd()
|
---|
| 82 | : base() {
|
---|
[8612] | 83 | Name = ItemName;
|
---|
| 84 | Description = ItemDescription;
|
---|
| 85 |
|
---|
| 86 | this.scaleParameter = new HyperParameter<DoubleValue>("Scale", "The scale parameter of the rational quadratic covariance function with ARD.");
|
---|
| 87 | this.inverseLengthParameter = new HyperParameter<DoubleArray>("InverseLength", "The inverse length parameter for automatic relevance determination.");
|
---|
| 88 | this.shapeParameter = new HyperParameter<DoubleValue>("Shape", "The shape parameter (alpha) of the rational quadratic covariance function with ARD.");
|
---|
| 89 |
|
---|
| 90 | Parameters.Add(scaleParameter);
|
---|
| 91 | Parameters.Add(inverseLengthParameter);
|
---|
| 92 | Parameters.Add(shapeParameter);
|
---|
| 93 |
|
---|
| 94 | RegisterEvents();
|
---|
[8565] | 95 | }
|
---|
| 96 |
|
---|
| 97 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 98 | return new CovarianceRQArd(this, cloner);
|
---|
| 99 | }
|
---|
| 100 |
|
---|
[8612] | 101 | [StorableHook(HookType.AfterDeserialization)]
|
---|
| 102 | private void AfterDeserialization() {
|
---|
| 103 | RegisterEvents();
|
---|
| 104 | }
|
---|
| 105 |
|
---|
| 106 | private void RegisterEvents() {
|
---|
| 107 | Util.AttachValueChangeHandler<DoubleValue, double>(scaleParameter, () => { sf2 = scaleParameter.Value.Value; });
|
---|
| 108 | Util.AttachValueChangeHandler<DoubleValue, double>(shapeParameter, () => { shape = shapeParameter.Value.Value; });
|
---|
| 109 | Util.AttachArrayChangeHandler<DoubleArray, double>(inverseLengthParameter, () => { inverseLength = inverseLengthParameter.Value.ToArray(); });
|
---|
| 110 | }
|
---|
| 111 |
|
---|
[8565] | 112 | public int GetNumberOfParameters(int numberOfVariables) {
|
---|
[8612] | 113 | return
|
---|
| 114 | (scaleParameter.Fixed ? 0 : 1) +
|
---|
| 115 | (shapeParameter.Fixed ? 0 : 1) +
|
---|
| 116 | (inverseLengthParameter.Fixed ? 0 : numberOfVariables);
|
---|
[8565] | 117 | }
|
---|
| 118 |
|
---|
| 119 | public void SetParameter(double[] hyp) {
|
---|
[8612] | 120 | int i = 0;
|
---|
| 121 | if (!scaleParameter.Fixed) {
|
---|
| 122 | scaleParameter.SetValue(new DoubleValue(Math.Exp(2 * hyp[i])));
|
---|
| 123 | i++;
|
---|
| 124 | }
|
---|
| 125 | if (!shapeParameter.Fixed) {
|
---|
| 126 | shapeParameter.SetValue(new DoubleValue(Math.Exp(hyp[i])));
|
---|
| 127 | i++;
|
---|
| 128 | }
|
---|
| 129 | if (!inverseLengthParameter.Fixed) {
|
---|
| 130 | inverseLengthParameter.SetValue(new DoubleArray(hyp.Skip(i).Select(e => 1.0 / Math.Exp(e)).ToArray()));
|
---|
| 131 | i += hyp.Skip(i).Count();
|
---|
| 132 | }
|
---|
| 133 | if (hyp.Length != i) throw new ArgumentException("The length of the parameter vector does not match the number of free parameters for CovarianceRQArd", "hyp");
|
---|
[8565] | 134 | }
|
---|
| 135 |
|
---|
| 136 |
|
---|
| 137 | public double GetCovariance(double[,] x, int i, int j) {
|
---|
| 138 | double d = i == j
|
---|
| 139 | ? 0.0
|
---|
| 140 | : Util.SqrDist(x, i, j, inverseLength);
|
---|
[8612] | 141 | return sf2 * Math.Pow(1 + 0.5 * d / shape, -shape);
|
---|
[8565] | 142 | }
|
---|
| 143 |
|
---|
| 144 | public IEnumerable<double> GetGradient(double[,] x, int i, int j) {
|
---|
| 145 | double d = i == j
|
---|
| 146 | ? 0.0
|
---|
| 147 | : Util.SqrDist(x, i, j, inverseLength);
|
---|
[8612] | 148 | double b = 1 + 0.5 * d / shape;
|
---|
[8565] | 149 | for (int k = 0; k < inverseLength.Length; k++) {
|
---|
[8612] | 150 | yield return sf2 * Math.Pow(b, -shape - 1) * Util.SqrDist(x[i, k] * inverseLength[k], x[j, k] * inverseLength[k]);
|
---|
[8565] | 151 | }
|
---|
[8612] | 152 | yield return 2 * sf2 * Math.Pow(b, -shape);
|
---|
| 153 | yield return sf2 * Math.Pow(b, -shape) * (0.5 * d / b - shape * Math.Log(b));
|
---|
[8565] | 154 | }
|
---|
| 155 |
|
---|
| 156 | public double GetCrossCovariance(double[,] x, double[,] xt, int i, int j) {
|
---|
| 157 | double d = Util.SqrDist(x, i, xt, j, inverseLength);
|
---|
[8612] | 158 | return sf2 * Math.Pow(1 + 0.5 * d / shape, -shape);
|
---|
[8565] | 159 | }
|
---|
| 160 | }
|
---|
| 161 | }
|
---|