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 HeuristicLab.Common;
|
---|
23 | using HeuristicLab.Core;
|
---|
24 | using HeuristicLab.Data;
|
---|
25 | using HeuristicLab.Operators;
|
---|
26 | using HeuristicLab.Optimization;
|
---|
27 | using HeuristicLab.Parameters;
|
---|
28 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
29 | using System;
|
---|
30 | using System.Linq;
|
---|
31 |
|
---|
32 | namespace HeuristicLab.Encodings.RealVectorEncoding {
|
---|
33 | [Item("CMAInitializer", "Initializes the covariance matrix and step size variables.")]
|
---|
34 | [StorableClass]
|
---|
35 | public class CMAInitializer : SingleSuccessorOperator, IRealVectorOperator, ICMAESInitializer, IIterationBasedOperator {
|
---|
36 |
|
---|
37 | public Type CMAType {
|
---|
38 | get { return typeof(CMAParameters); }
|
---|
39 | }
|
---|
40 |
|
---|
41 | #region Parameter Properties
|
---|
42 | public IValueLookupParameter<IntValue> DimensionParameter {
|
---|
43 | get { return (IValueLookupParameter<IntValue>)Parameters["Dimension"]; }
|
---|
44 | }
|
---|
45 |
|
---|
46 | public IValueLookupParameter<DoubleArray> InitialSigmaParameter {
|
---|
47 | get { return (IValueLookupParameter<DoubleArray>)Parameters["InitialSigma"]; }
|
---|
48 | }
|
---|
49 |
|
---|
50 | public IValueLookupParameter<DoubleArray> MaxSigmaParameter {
|
---|
51 | get { return (IValueLookupParameter<DoubleArray>)Parameters["MaxSigma"]; }
|
---|
52 | }
|
---|
53 |
|
---|
54 | public IValueLookupParameter<DoubleArray> MinSigmaParameter {
|
---|
55 | get { return (IValueLookupParameter<DoubleArray>)Parameters["MinSigma"]; }
|
---|
56 | }
|
---|
57 |
|
---|
58 | public ILookupParameter<IntValue> IterationsParameter {
|
---|
59 | get { return (ILookupParameter<IntValue>)Parameters["Iterations"]; }
|
---|
60 | }
|
---|
61 |
|
---|
62 | public IValueLookupParameter<IntValue> MaximumIterationsParameter {
|
---|
63 | get { return (IValueLookupParameter<IntValue>)Parameters["MaximumIterations"]; }
|
---|
64 | }
|
---|
65 |
|
---|
66 | public IValueLookupParameter<IntValue> InitialIterationsParameter {
|
---|
67 | get { return (IValueLookupParameter<IntValue>)Parameters["InitialIterations"]; }
|
---|
68 | }
|
---|
69 |
|
---|
70 | public ILookupParameter<IntValue> PopulationSizeParameter {
|
---|
71 | get { return (ILookupParameter<IntValue>)Parameters["PopulationSize"]; }
|
---|
72 | }
|
---|
73 |
|
---|
74 | public ILookupParameter<IntValue> MuParameter {
|
---|
75 | get { return (ILookupParameter<IntValue>)Parameters["Mu"]; }
|
---|
76 | }
|
---|
77 |
|
---|
78 | public ILookupParameter<CMAParameters> StrategyParametersParameter {
|
---|
79 | get { return (ILookupParameter<CMAParameters>)Parameters["StrategyParameters"]; }
|
---|
80 | }
|
---|
81 | #endregion
|
---|
82 |
|
---|
83 | [StorableConstructor]
|
---|
84 | protected CMAInitializer(bool deserializing) : base(deserializing) { }
|
---|
85 | protected CMAInitializer(CMAInitializer original, Cloner cloner) : base(original, cloner) { }
|
---|
86 | public CMAInitializer()
|
---|
87 | : base() {
|
---|
88 | Parameters.Add(new ValueLookupParameter<IntValue>("Dimension", "The problem dimension (N)."));
|
---|
89 | Parameters.Add(new ValueLookupParameter<DoubleArray>("InitialSigma", "The initial value for Sigma (need to be > 0), can be single dimensioned or an array that should be equal to the size of the vector.", new DoubleArray(new[] { 0.5 })));
|
---|
90 | Parameters.Add(new ValueLookupParameter<DoubleArray>("MaxSigma", "The maximum sigma value can be omitted, given as one value for all dimensions or a value for each dimension."));
|
---|
91 | Parameters.Add(new ValueLookupParameter<DoubleArray>("MinSigma", "The minimum sigma value can be omitted, given as one value for all dimensions or a value for each dimension."));
|
---|
92 | Parameters.Add(new LookupParameter<IntValue>("Iterations", "The current iteration that is being processed."));
|
---|
93 | Parameters.Add(new ValueLookupParameter<IntValue>("MaximumIterations", "The maximum number of iterations to be processed."));
|
---|
94 | Parameters.Add(new ValueLookupParameter<IntValue>("InitialIterations", "The number of iterations that should be performed using the diagonal covariance matrix only."));
|
---|
95 | Parameters.Add(new LookupParameter<IntValue>("PopulationSize", "The population size (lambda)."));
|
---|
96 | Parameters.Add(new LookupParameter<IntValue>("Mu", "The number of offspring considered for updating of the strategy parameters."));
|
---|
97 | Parameters.Add(new LookupParameter<CMAParameters>("StrategyParameters", "The strategy parameters for real-encoded CMA-ES."));
|
---|
98 | }
|
---|
99 |
|
---|
100 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
101 | return new CMAInitializer(this, cloner);
|
---|
102 | }
|
---|
103 |
|
---|
104 | public override IOperation Apply() {
|
---|
105 | var N = DimensionParameter.ActualValue.Value;
|
---|
106 | var lambda = PopulationSizeParameter.ActualValue.Value;
|
---|
107 |
|
---|
108 | var sp = new CMAParameters(N, lambda);
|
---|
109 |
|
---|
110 | var s = InitialSigmaParameter.ActualValue;
|
---|
111 | if (s == null || s.Length == 0) throw new InvalidOperationException("Initial standard deviation (sigma) must be given.");
|
---|
112 | var sigma = s.Max();
|
---|
113 | if (sigma <= 0) throw new InvalidOperationException("Initial standard deviation (sigma) must be > 0.");
|
---|
114 |
|
---|
115 | var pc = new double[N]; // evolution paths for C
|
---|
116 | var ps = new double[N]; // evolution paths for sigma
|
---|
117 | var B = new double[N, N]; // B defines the coordinate system
|
---|
118 | var D = new double[N]; // diagonal D defines the scaling
|
---|
119 | var C = new double[N, N]; // covariance matrix C
|
---|
120 | var BDz = new double[N];
|
---|
121 | double minSqrtdiagC = int.MaxValue, maxSqrtdiagC = int.MinValue;
|
---|
122 | for (int i = 0; i < N; i++) {
|
---|
123 | B[i, i] = 1;
|
---|
124 | if (s.Length == 1) D[i] = 1;
|
---|
125 | else if (s.Length == N) D[i] = s[i] / sigma;
|
---|
126 | else throw new InvalidOperationException("Initial standard deviation (sigma) must either contain only one value for all dimension or for every dimension.");
|
---|
127 | if (D[i] <= 0) throw new InvalidOperationException("Initial standard deviation (sigma) values must all be > 0.");
|
---|
128 | C[i, i] = D[i] * D[i];
|
---|
129 | if (Math.Sqrt(C[i, i]) < minSqrtdiagC) minSqrtdiagC = Math.Sqrt(C[i, i]);
|
---|
130 | if (Math.Sqrt(C[i, i]) > maxSqrtdiagC) maxSqrtdiagC = Math.Sqrt(C[i, i]);
|
---|
131 | }
|
---|
132 |
|
---|
133 | // ensure maximal and minimal standard deviations
|
---|
134 | var minSigma = MinSigmaParameter.ActualValue;
|
---|
135 | if (minSigma != null && minSigma.Length > 0) {
|
---|
136 | for (int i = 0; i < N; i++) {
|
---|
137 | var d = minSigma[(int)Math.Min(i, minSigma.Length - 1)];
|
---|
138 | if (d > sigma * minSqrtdiagC) sigma = d / minSqrtdiagC;
|
---|
139 | }
|
---|
140 | }
|
---|
141 | var maxSigma = MaxSigmaParameter.ActualValue;
|
---|
142 | if (maxSigma != null && maxSigma.Length > 0) {
|
---|
143 | for (int i = 0; i < N; i++) {
|
---|
144 | var d = maxSigma[(int)Math.Min(i, maxSigma.Length - 1)];
|
---|
145 | if (d > sigma * maxSqrtdiagC) sigma = d / maxSqrtdiagC;
|
---|
146 | }
|
---|
147 | }
|
---|
148 | // end ensure ...
|
---|
149 |
|
---|
150 | // testAndCorrectNumerics
|
---|
151 | double fac = 1;
|
---|
152 | if (D.Max() < 1e-6)
|
---|
153 | fac = 1.0 / D.Max();
|
---|
154 | else if (D.Min() > 1e4)
|
---|
155 | fac = 1.0 / D.Min();
|
---|
156 |
|
---|
157 | if (fac != 1.0) {
|
---|
158 | sigma /= fac;
|
---|
159 | for (int i = 0; i < N; i++) {
|
---|
160 | pc[i] *= fac;
|
---|
161 | D[i] *= fac;
|
---|
162 | for (int j = 0; j < N; j++)
|
---|
163 | C[i, j] *= fac * fac;
|
---|
164 | }
|
---|
165 | }
|
---|
166 | // end testAndCorrectNumerics
|
---|
167 |
|
---|
168 | var initialIterations = InitialIterationsParameter.ActualValue;
|
---|
169 | if (initialIterations == null) {
|
---|
170 | initialIterations = new IntValue(0);
|
---|
171 | }
|
---|
172 |
|
---|
173 | double maxD = D.Max(), minD = D.Min();
|
---|
174 | if (minD == 0) sp.AxisRatio = new DoubleValue(double.PositiveInfinity);
|
---|
175 | else sp.AxisRatio = new DoubleValue(maxD / minD);
|
---|
176 | sp.PC = new DoubleArray(pc);
|
---|
177 | sp.PS = new DoubleArray(ps);
|
---|
178 | sp.B = new DoubleMatrix(B);
|
---|
179 | sp.D = new DoubleArray(D);
|
---|
180 | sp.C = new DoubleMatrix(C);
|
---|
181 | sp.BDz = new DoubleArray(BDz);
|
---|
182 | sp.Sigma = new DoubleValue(sigma);
|
---|
183 | if (maxSigma != null) sp.MaxSigma = (DoubleArray)maxSigma.Clone();
|
---|
184 | if (minSigma != null) sp.MinSigma = (DoubleArray)minSigma.Clone();
|
---|
185 | sp.InitialIterations = new IntValue(initialIterations.Value);
|
---|
186 |
|
---|
187 | MuParameter.ActualValue = new IntValue(sp.Mu.Value);
|
---|
188 | StrategyParametersParameter.ActualValue = sp;
|
---|
189 | return base.Apply();
|
---|
190 | }
|
---|
191 | }
|
---|
192 | } |
---|