1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 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 HeuristicLab.Common;
|
---|
25 | using HeuristicLab.Core;
|
---|
26 | using HeuristicLab.Data;
|
---|
27 | using HeuristicLab.Parameters;
|
---|
28 | using HEAL.Attic;
|
---|
29 |
|
---|
30 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
31 | [StorableType("F60E0A63-0107-44E3-920B-BB5B09E9DDDF")]
|
---|
32 | [Item(Name = "CovarianceNeuralNetwork",
|
---|
33 | Description = "Neural network covariance function for Gaussian processes.")]
|
---|
34 | public sealed class CovarianceNeuralNetwork : ParameterizedNamedItem, ICovarianceFunction {
|
---|
35 | public IValueParameter<DoubleValue> ScaleParameter {
|
---|
36 | get { return (IValueParameter<DoubleValue>)Parameters["Scale"]; }
|
---|
37 | }
|
---|
38 |
|
---|
39 | public IValueParameter<DoubleValue> LengthParameter {
|
---|
40 | get { return (IValueParameter<DoubleValue>)Parameters["Length"]; }
|
---|
41 | }
|
---|
42 | private bool HasFixedScaleParameter {
|
---|
43 | get { return ScaleParameter.Value != null; }
|
---|
44 | }
|
---|
45 | private bool HasFixedLengthParameter {
|
---|
46 | get { return LengthParameter.Value != null; }
|
---|
47 | }
|
---|
48 |
|
---|
49 | [StorableConstructor]
|
---|
50 | private CovarianceNeuralNetwork(StorableConstructorFlag _) : base(_) {
|
---|
51 | }
|
---|
52 |
|
---|
53 | private CovarianceNeuralNetwork(CovarianceNeuralNetwork original, Cloner cloner)
|
---|
54 | : base(original, cloner) {
|
---|
55 | }
|
---|
56 |
|
---|
57 | public CovarianceNeuralNetwork()
|
---|
58 | : base() {
|
---|
59 | Name = ItemName;
|
---|
60 | Description = ItemDescription;
|
---|
61 |
|
---|
62 | Parameters.Add(new OptionalValueParameter<DoubleValue>("Scale", "The scale parameter."));
|
---|
63 | Parameters.Add(new OptionalValueParameter<DoubleValue>("Length", "The length parameter."));
|
---|
64 | }
|
---|
65 |
|
---|
66 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
67 | return new CovarianceNeuralNetwork(this, cloner);
|
---|
68 | }
|
---|
69 |
|
---|
70 | public int GetNumberOfParameters(int numberOfVariables) {
|
---|
71 | return
|
---|
72 | (HasFixedScaleParameter ? 0 : 1) +
|
---|
73 | (HasFixedLengthParameter ? 0 : 1);
|
---|
74 | }
|
---|
75 |
|
---|
76 | public void SetParameter(double[] p) {
|
---|
77 | double scale, length;
|
---|
78 | GetParameterValues(p, out scale, out length);
|
---|
79 | ScaleParameter.Value = new DoubleValue(scale);
|
---|
80 | LengthParameter.Value = new DoubleValue(length);
|
---|
81 | }
|
---|
82 |
|
---|
83 |
|
---|
84 | private void GetParameterValues(double[] p, out double scale, out double length) {
|
---|
85 | // gather parameter values
|
---|
86 | int c = 0;
|
---|
87 | if (HasFixedLengthParameter) {
|
---|
88 | length = LengthParameter.Value.Value;
|
---|
89 | } else {
|
---|
90 | length = Math.Exp(2 * p[c]);
|
---|
91 | c++;
|
---|
92 | }
|
---|
93 |
|
---|
94 | if (HasFixedScaleParameter) {
|
---|
95 | scale = ScaleParameter.Value.Value;
|
---|
96 | } else {
|
---|
97 | scale = Math.Exp(2 * p[c]);
|
---|
98 | c++;
|
---|
99 | }
|
---|
100 | if (p.Length != c) throw new ArgumentException("The length of the parameter vector does not match the number of free parameters for CovarianceNeuralNetwork", "p");
|
---|
101 | }
|
---|
102 |
|
---|
103 | public ParameterizedCovarianceFunction GetParameterizedCovarianceFunction(double[] p, int[] columnIndices) {
|
---|
104 | double length, scale;
|
---|
105 | GetParameterValues(p, out scale, out length);
|
---|
106 | var fixedLength = HasFixedLengthParameter;
|
---|
107 | var fixedScale = HasFixedScaleParameter;
|
---|
108 |
|
---|
109 | var cov = new ParameterizedCovarianceFunction();
|
---|
110 | cov.Covariance = (x, i, j) => {
|
---|
111 | double sx = 1.0;
|
---|
112 | double s1 = 1.0;
|
---|
113 | double s2 = 1.0;
|
---|
114 | for (int c = 0; c < columnIndices.Length; c++) {
|
---|
115 | var col = columnIndices[c];
|
---|
116 | sx += x[i, col] * x[j, col];
|
---|
117 | s1 += x[i, col] * x[i, col];
|
---|
118 | s2 += x[j, col] * x[j, col];
|
---|
119 | }
|
---|
120 |
|
---|
121 | return (scale * Math.Asin(sx / (Math.Sqrt((length + s1) * (length + s2)))));
|
---|
122 | };
|
---|
123 | cov.CrossCovariance = (x, xt, i, j) => {
|
---|
124 | double sx = 1.0;
|
---|
125 | double s1 = 1.0;
|
---|
126 | double s2 = 1.0;
|
---|
127 | for (int c = 0; c < columnIndices.Length; c++) {
|
---|
128 | var col = columnIndices[c];
|
---|
129 | sx += x[i, col] * xt[j, col];
|
---|
130 | s1 += x[i, col] * x[i, col];
|
---|
131 | s2 += xt[j, col] * xt[j, col];
|
---|
132 | }
|
---|
133 |
|
---|
134 | return (scale * Math.Asin(sx / (Math.Sqrt((length + s1) * (length + s2)))));
|
---|
135 | };
|
---|
136 | cov.CovarianceGradient = (x, i, j) => GetGradient(x, i, j, length, scale, columnIndices, fixedLength, fixedScale);
|
---|
137 | return cov;
|
---|
138 | }
|
---|
139 |
|
---|
140 | // order of returned gradients must match the order in GetParameterValues!
|
---|
141 | private static IList<double> GetGradient(double[,] x, int i, int j, double length, double scale, int[] columnIndices,
|
---|
142 | bool fixedLength, bool fixedScale) {
|
---|
143 | double sx = 1.0;
|
---|
144 | double s1 = 1.0;
|
---|
145 | double s2 = 1.0;
|
---|
146 | for (int c = 0; c < columnIndices.Length; c++) {
|
---|
147 | var col = columnIndices[c];
|
---|
148 | sx += x[i, col] * x[j, col];
|
---|
149 | s1 += x[i, col] * x[i, col];
|
---|
150 | s2 += x[j, col] * x[j, col];
|
---|
151 | }
|
---|
152 | var h = (length + s1) * (length + s2);
|
---|
153 | var f = sx / Math.Sqrt(h);
|
---|
154 |
|
---|
155 | var g = new List<double>(2);
|
---|
156 | if (!fixedLength) g.Add(-scale / Math.Sqrt(1.0 - f * f) * ((length * sx * (2.0 * length + s1 + s2)) / Math.Pow(h, 3.0 / 2.0)));
|
---|
157 | if (!fixedScale) g.Add(2.0 * scale * Math.Asin(f));
|
---|
158 | return g;
|
---|
159 | }
|
---|
160 | }
|
---|
161 | }
|
---|