1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2013 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 System.Linq.Expressions;
|
---|
26 | using AutoDiff;
|
---|
27 | using HeuristicLab.Common;
|
---|
28 | using HeuristicLab.Core;
|
---|
29 | using HeuristicLab.Data;
|
---|
30 | using HeuristicLab.Parameters;
|
---|
31 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
32 |
|
---|
33 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
34 | [StorableClass]
|
---|
35 | [Item(Name = "CovarianceNeuralNetwork",
|
---|
36 | Description = "Neural network covariance function for Gaussian processes.")]
|
---|
37 | public sealed class CovarianceNeuralNetwork : ParameterizedNamedItem, ICovarianceFunction {
|
---|
38 | public IValueParameter<DoubleValue> ScaleParameter {
|
---|
39 | get { return (IValueParameter<DoubleValue>)Parameters["Scale"]; }
|
---|
40 | }
|
---|
41 |
|
---|
42 | public IValueParameter<DoubleValue> LengthParameter {
|
---|
43 | get { return (IValueParameter<DoubleValue>)Parameters["Length"]; }
|
---|
44 | }
|
---|
45 |
|
---|
46 | [StorableConstructor]
|
---|
47 | private CovarianceNeuralNetwork(bool deserializing)
|
---|
48 | : base(deserializing) {
|
---|
49 | }
|
---|
50 |
|
---|
51 | private CovarianceNeuralNetwork(CovarianceNeuralNetwork original, Cloner cloner)
|
---|
52 | : base(original, cloner) {
|
---|
53 | }
|
---|
54 |
|
---|
55 | public CovarianceNeuralNetwork()
|
---|
56 | : base() {
|
---|
57 | Name = ItemName;
|
---|
58 | Description = ItemDescription;
|
---|
59 |
|
---|
60 | Parameters.Add(new OptionalValueParameter<DoubleValue>("Scale", "The scale parameter."));
|
---|
61 | Parameters.Add(new OptionalValueParameter<DoubleValue>("Length", "The length parameter."));
|
---|
62 | }
|
---|
63 |
|
---|
64 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
65 | return new CovarianceNeuralNetwork(this, cloner);
|
---|
66 | }
|
---|
67 |
|
---|
68 | public int GetNumberOfParameters(int numberOfVariables) {
|
---|
69 | return
|
---|
70 | (ScaleParameter.Value != null ? 0 : 1) +
|
---|
71 | (LengthParameter.Value != null ? 0 : 1);
|
---|
72 | }
|
---|
73 |
|
---|
74 | public void SetParameter(double[] p) {
|
---|
75 | double scale, length;
|
---|
76 | GetParameterValues(p, out scale, out length);
|
---|
77 | ScaleParameter.Value = new DoubleValue(scale);
|
---|
78 | LengthParameter.Value = new DoubleValue(length);
|
---|
79 | }
|
---|
80 |
|
---|
81 |
|
---|
82 | private void GetParameterValues(double[] p, out double scale, out double length) {
|
---|
83 | // gather parameter values
|
---|
84 | int c = 0;
|
---|
85 | if (LengthParameter.Value != null) {
|
---|
86 | length = LengthParameter.Value.Value;
|
---|
87 | } else {
|
---|
88 | length = Math.Exp(2 * p[c]);
|
---|
89 | c++;
|
---|
90 | }
|
---|
91 |
|
---|
92 | if (ScaleParameter.Value != null) {
|
---|
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 CovarianceNeuralNetwork", "p");
|
---|
99 | }
|
---|
100 |
|
---|
101 |
|
---|
102 | private static Func<Term, UnaryFunc> asin = UnaryFunc.Factory(
|
---|
103 | x => Math.Asin(x), // evaluate
|
---|
104 | x => 1 / Math.Sqrt(1 - x * x)); // derivative of atan
|
---|
105 | private static Func<Term, UnaryFunc> sqrt = UnaryFunc.Factory(
|
---|
106 | x => Math.Sqrt(x),
|
---|
107 | x => 1 / (2 * Math.Sqrt(x)));
|
---|
108 |
|
---|
109 | public ParameterizedCovarianceFunction GetParameterizedCovarianceFunction(double[] p, IEnumerable<int> columnIndices) {
|
---|
110 | double length, scale;
|
---|
111 | GetParameterValues(p, out scale, out length);
|
---|
112 | // create functions
|
---|
113 | AutoDiff.Variable p0 = new AutoDiff.Variable();
|
---|
114 | AutoDiff.Variable p1 = new AutoDiff.Variable();
|
---|
115 | var l = TermBuilder.Exp(2.0 * p0);
|
---|
116 | var s = TermBuilder.Exp(2.0 * p1);
|
---|
117 | AutoDiff.Variable[] x1 = new AutoDiff.Variable[columnIndices.Count()];
|
---|
118 | AutoDiff.Variable[] x2 = new AutoDiff.Variable[columnIndices.Count()];
|
---|
119 | AutoDiff.Term sx = 1;
|
---|
120 | AutoDiff.Term s1 = 1;
|
---|
121 | AutoDiff.Term s2 = 1;
|
---|
122 | foreach (var k in columnIndices) {
|
---|
123 | x1[k] = new AutoDiff.Variable();
|
---|
124 | x2[k] = new AutoDiff.Variable();
|
---|
125 | sx += x1[k] * x2[k];
|
---|
126 | s1 += x1[k] * x1[k];
|
---|
127 | s2 += x2[k] * x2[k];
|
---|
128 | }
|
---|
129 |
|
---|
130 | var parameter = x1.Concat(x2).Concat(new AutoDiff.Variable[] { p0, p1 }).ToArray();
|
---|
131 | var values = new double[x1.Length + x2.Length + 2];
|
---|
132 | var c = (s * asin(sx / (sqrt((l + s1) * (l + s2))))).Compile(parameter);
|
---|
133 |
|
---|
134 | var cov = new ParameterizedCovarianceFunction();
|
---|
135 | cov.Covariance = (x, i, j) => {
|
---|
136 | int k = 0;
|
---|
137 | foreach (var col in columnIndices) {
|
---|
138 | values[k] = x[i, col];
|
---|
139 | k++;
|
---|
140 | }
|
---|
141 | foreach (var col in columnIndices) {
|
---|
142 | values[k] = x[j, col];
|
---|
143 | k++;
|
---|
144 | }
|
---|
145 | values[k] = Math.Log(Math.Sqrt(length));
|
---|
146 | values[k + 1] = Math.Log(Math.Sqrt(scale));
|
---|
147 | return c.Evaluate(values);
|
---|
148 | };
|
---|
149 | cov.CrossCovariance = (x, xt, i, j) => {
|
---|
150 | int k = 0;
|
---|
151 | foreach (var col in columnIndices) {
|
---|
152 | values[k] = x[i, col];
|
---|
153 | k++;
|
---|
154 | }
|
---|
155 | foreach (var col in columnIndices) {
|
---|
156 | values[k] = xt[j, col];
|
---|
157 | k++;
|
---|
158 | }
|
---|
159 | values[k] = Math.Log(Math.Sqrt(length));
|
---|
160 | values[k + 1] = Math.Log(Math.Sqrt(scale));
|
---|
161 | return c.Evaluate(values);
|
---|
162 | };
|
---|
163 | cov.CovarianceGradient = (x, i, j) => {
|
---|
164 | int k = 0;
|
---|
165 | foreach (var col in columnIndices) {
|
---|
166 | values[k] = x[i, col];
|
---|
167 | k++;
|
---|
168 | }
|
---|
169 | foreach (var col in columnIndices) {
|
---|
170 | values[k] = x[j, col];
|
---|
171 | k++;
|
---|
172 | }
|
---|
173 | values[k] = Math.Log(Math.Sqrt(length));
|
---|
174 | values[k + 1] = Math.Log(Math.Sqrt(scale));
|
---|
175 | return c.Differentiate(values).Item1.Skip(columnIndices.Count() * 2);
|
---|
176 | };
|
---|
177 | return cov;
|
---|
178 | }
|
---|
179 |
|
---|
180 | }
|
---|
181 | }
|
---|