17-09-2012, 05:20 PM
Evolutionary Multiobjective Optimization using a Cultural Algorithm
1Evolutionary Multiobjective.pdf (Size: 149.42 KB / Downloads: 85)
Abstract
In this paper, we present the first proposal to use a
cultural algorithm to solve multiobjective optimization problems.
Our proposal uses evolutionary programming, Pareto ranking and
elitism (i.e., an external population). The approach proposed is validated
using several examples taken from the specialized literature.
Our results are compared with respect to the NSGA-II, which is
an algorithm representative of the state-of-the-art in evolutionary
multiobjective optimization. The performance of our approach indicates
that cultural algorithms are a viable alternative for multiobjective
optimization.
INTRODUCTION
In recent years, evolutionary algorithms have been
widely used for multiobjective optimization, obtaining
very promising results [3], [2]. An important advantage
of evolutionary algorithms over traditional techniques
used for multiobjective optimization is that the former
operate over a set of solutions at a time. Therefore, evolutionary
algorithms can produce several elements of the
Pareto optimal set in a single run rather than generating
one at a time as traditional mathematical programming
approaches.
Cultural algorithms [18] are a technique that incorporates
domain knowledge obtained during the evolutionary
process as to make the search process more efficient. Cultural
algorithms have been successfully applied to several
types of optimization problems (e.g., in constrained
single-objective optimization problems [20], [12], [13],
[4]). However, until now, nobody had proposed a cultural
algorithm for multiobjective optimization adopting
Pareto ranking and elitism. Ours is then the first proposal
in this direction.
NOTIONS OF CULTURAL ALGORITHMS
Cultural algorithms were developed by Robert G.
Reynolds as a complement to the metaphor adopted by
evolutionary algorithms, which was mainly focused on
genetic concepts and on the natural selection mechanism
[18]. Cultural algorithms are based on some theories proposed
in sociology and archaelogy to model cultural evolution.
Such theories indicate that cultural evolution can
be seen as an inheritance process that occurs at two levels:
the micro-evolutionary level, and the macro-evolutionary
level.
CONCLUSIONS AND FUTURE WORK
In this paper we have introduced the first proposal to
use cultural algorithms for multiobjective optimization
incorporating Pareto-based selection. We have shown
how adding a belief space to an evolutionary programming
algorithm can be beneficial and represents a viable
alternative to solve multiobjective problems.
The small comparative study presented in this paper
validates the viability of our proposal. However, we obviously
still need to do a more in-depth study in which other
evolutionary multiobjective optimization techniques and
more test functions are involved. Additionally, there are
still a few limitations of CAEP that require a more careful
analysis. For example, CAEP tends to lose diversity very
quickly in some cases.