ReferenceVectorAdaptation.h
Go to the documentation of this file.
1//===========================================================================
2/*!
3 *
4 *
5 * \brief Reference vector adaptation for the RVEA algorithm.
6 *
7 * \author Bjoern Bugge Grathwohl
8 * \date March 2017
9 *
10 * \par Copyright 1995-2017 Shark Development Team
11 *
12 * <BR><HR>
13 * This file is part of Shark.
14 * <https://shark-ml.github.io/Shark/>
15 *
16 * Shark is free software: you can redistribute it and/or modify
17 * it under the terms of the GNU Lesser General Public License as published
18 * by the Free Software Foundation, either version 3 of the License, or
19 * (at your option) any later version.
20 *
21 * Shark is distributed in the hope that it will be useful,
22 * but WITHOUT ANY WARRANTY; without even the implied warranty of
23 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24 * GNU Lesser General Public License for more details.
25 *
26 * You should have received a copy of the GNU Lesser General Public License
27 * along with Shark. If not, see <http://www.gnu.org/licenses/>.
28 *
29 */
30//===========================================================================
31
32#ifndef SHARK_ALGORITHMS_DIRECT_SEARCH_OPERATORS_REFERENCE_VECTOR_ADAPTATION_H
33#define SHARK_ALGORITHMS_DIRECT_SEARCH_OPERATORS_REFERENCE_VECTOR_ADAPTATION_H
34
35#include <shark/LinAlg/Base.h>
36#include <shark/Algorithms/DirectSearch/Individual.h>
37
38namespace shark {
39/**
40 * \brief Reference vector adaptation for the RVEA algorithm.
41 *
42 * This operator is supposed to be applied regularly in the RVEA algorithm to
43 * adjust the set of reference vectors to better match a scaled pareto front.
44 * See below paper for details:
45 * R. Cheng, Y. Jin, M. Olhofer, and B. Sendhoff, “A reference vector guided
46 * evolutionary algorithm for many-objective optimization,” IEEE Transactions on
47 * Evolutionary Computation, Vol 20, Issue 5, October 2016
48 * http://dx.doi.org/10.1109/TEVC.2016.2519378
49 */
50template <typename IndividualType>
51struct ReferenceVectorAdaptation
52{
53 /**
54 * \brief Apply adaptation operator and update the angles.
55 */
56 void operator()(
57 std::vector<IndividualType> const & population,
58 RealMatrix & referenceVectors,
59 RealVector & minAngles)
60 {
61 auto f = unpenalizedFitness(population);
62 const std::size_t sz = f.size();
63 const std::size_t w = f[0].size();
64 RealVector diff(w);
65 for(std::size_t i = 0; i < w; ++i)
66 {
67 double max = std::numeric_limits<double>::min();
68 double min = std::numeric_limits<double>::max();
69 for(std::size_t j = 0; j < sz; ++j)
70 {
71 max = std::max(max, f[j][i]);
72 min = std::min(min, f[j][i]);
73 }
74 double d = max - min;
75 diff[i] = (d == 0) ? 1 : d;
76 }
77 referenceVectors = m_initVecs * repeat(diff, m_initVecs.size1());
78 for(std::size_t i = 0; i < referenceVectors.size1(); ++i)
79 {
80 row(referenceVectors, i) /= norm_2(row(referenceVectors, i));
81 }
82 updateAngles(referenceVectors, minAngles);
83 }
84
85 /**
86 * \brief Compute the minimum angles between unit vectors.
87 */
88 static void updateAngles(
89 RealMatrix const & referenceVectors,
90 RealVector & minAngles)
91 {
92 const std::size_t s = referenceVectors.size1();
93 const RealMatrix m = acos(prod(referenceVectors,
94 trans(referenceVectors))) +
95 to_diagonal(RealVector(s, 1e10));
96 for(std::size_t i = 0; i < s; ++i)
97 {
98 minAngles[i] = min(row(m, i));
99 }
100 }
101
102 template <typename Archive>
103 void serialize(Archive & archive)
104 {
105 archive & m_initVecs;
106 }
107
108 /// \brief The set of initial reference vectors.
109 ///
110 /// This must be set before the operator is called the first time.
111 RealMatrix m_initVecs;
112};
113
114} // namespace shark
115
116#endif // SHARK_ALGORITHMS_DIRECT_SEARCH_OPERATORS_REFERENCE_VECTOR_ADAPTATION_H