Guest Editorial Special Issue on Memetic Algorithms - IEEE Xplore

14 downloads 104836 Views 89KB Size Report
and floor-planning problems, to pattern recognition, vehicle routing, control ... classes, thus rendering the algorithm very general and scalable. The second paper, by .... California, San Diego, CA, 1994. Yew-Soon Ong ... Hisao Ishibuchi (M'93) received the B.S. and M.S. degrees in precision mechanics from Kyoto. University ...
2

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 37, NO. 1, FEBRUARY 2007

Guest Editorial Special Issue on Memetic Algorithms I. I NTRODUCTION

T

HE USE of sophisticated machine intelligence approaches for solving complex problems in science and engineering has increased steadily over the last 20 years. Within this growing trend, which relies heavily on state-of-the-art optimization and design strategies, the methodology known as Memetic Algorithms is, perhaps, one of the most successful stories. Memetic algorithms [1]–[3] are population-based metaheuristic search methods that are inspired by Darwinian principles of natural selection and Dawkins’ notion of a meme defined as a unit of cultural evolution that is capable of local refinements. The metaphorical parallels to, on the one hand, Darwinian evolution and, on the other hand, between memes and domainspecific (local search) heuristics are captured within memetic algorithms, thus rendering a methodology that balances well generality and problem specificity. In diverse contexts, memetic algorithms have also been used under the name of hybrid evolutionary algorithms, Baldwinian evolutionary algorithms, Lamarkian evolutionary algorithms, cultural algorithms, or genetic local search. Many different instantiations of memetic algorithms have been reported across a wide variety of application domains that range from scheduling and floor-planning problems, to pattern recognition, vehicle routing, control systems, aircraft, and drug design, to name but a few. This large body of evidence has revealed that memetic algorithms not only converge to high-quality solutions, but also search vast, and sometimes noisy, solution spaces more efficiently than their conventional counterparts. Thus, memetic algorithms are the preferred methodology for many real-world applications. In response to this “Memetic Algorithms” special issue’s call for papers, we received an unprecedented number of submissions, 54 in total, reflecting the vigor and health of this research field. Each of this submissions was rigorously reviewed by at least three independent reviewers. Given the volume of papers received, we were able to accept only the top ten most outstanding submissions. Thus, many papers that did not make it into the special issue were still of very good quality. The papers that appear in this special issue reflect the state-ofthe-art in research, development, and application of memetic algorithms to a wide range of relevant topics and applications. II. S TRUCTURE OF THE S PECIAL I SSUE This papers presented in this special issue are loosely grouped into the following two categories: memetic algorithm

Digital Object Identifier 10.1109/TSMCB.2006.883274

methodologies and domain-specific memetic algorithms. The former category focuses primarily on new design methodologies of memetic algorithms while the latter concentrates on specialized memetic algorithms designed for tackling domainspecified applications. A. Memetic Algorithm Methodologies The papers by Smith, Liu et al., and Caponio et al. represent well the latest trends in adaptive and self-adaptive optimization techniques using the terminology provided in [1]. The first paper, “Coevolving Memetic Algorithms: A Review and Progress Report” by Smith, describes a self-adaptive memetic algorithm framework, named COMA, for creating new and robust scalable optimization techniques using the concept of coevolving memes. These coevolving memes encode on-the-fly alternative definitions of local search operators. The framework encompasses a general structure for evolving a population of local search mechanisms in tandem with the solutions to which they are to be applied. It is shown that COMA is capable of discovering and exploiting the structures and regularities of the search space for a variety of optimization problem classes, thus rendering the algorithm very general and scalable. The second paper, by Liu et al., “An Effective PSO-Based Memetic Algorithm for Flow Shop Scheduling,” and the third paper, by Caponio et al., “A Fast Adaptive Memetic Algorithm for Online and Offline Control Design of PMSM Drives,” consider the design of adaptive memetic algorithms for solving real-world applications. Both papers adapt the memes or local heuristics used to locally improve the evolved solutions and balance the degree of exploration and exploitation throughout the search process. More specifically, Liu et al.’s paper tackles the permutation flow shop scheduling problem [3] using a particle swarm optimization-based adaptive memetic algorithm. Their algorithm employs a sophisticated combination of particle swarm optimization, adaptive local searchers, and meta-Lamarckian learning. Caponio et al.’s paper describes a fast genetics-based adaptive memetic algorithm for optimal control systems in a permanent magnet synchronous motor. This memetic algorithm uses a clever scheduling policy for its local search operators that can be used either in tandem or concurrently. The approach reports success even in the presence of noisy objective functions. The fourth paper, by Liu et al., “A Multiobjective Memetic Algorithm Based on Particle Swarm Optimization,” studies a multiobjective particle swarm optimization based on fuzzy particle update and synchronized local search for continuous optimization [4]. The authors propose specific mechanisms that help to overcome premature convergence and enhance diversity

1083-4419/$25.00 © 2007 IEEE

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 37, NO. 1, FEBRUARY 2007

bookkeeping within the swarm. The benefits of their improved method is then assessed against five existing state-of-the-art multiobjective optimization algorithms. In their paper, “On the Hybridization of Memetic Algorithms With Branch-and-Bound Techniques,” Gallardo et al. explore in detail the best ways to hybridize some branch-and-bound (BnB)-based operators and memetic algorithms in the context of combinatorial optimization problems [5]. The synergy between the memetic algorithm and the BnB arises due to the fact that the former provides improved bounds that the BnB algorithm can use to purge and further reduce the problem queue, while the latter biases the search that the memetic algorithm performs toward more promising regions of the search space. The paper shows that the proposed BnB-based memetic algorithm is able to efficiently achieve good solution quality on multidimensional 0–1 knapsack and shortest common supersequence optimization problems improving over the traditional counterparts (especially for large problem instances). B. Domain-Specific Memetic Algorithms Zhang et al.’s paper, “Evolutionary Algorithms Refining a Heuristic: A Hybrid Method for Shared-Path Protections in WDM Networks Under SRLG Constraints,” describes a threephase parameterized construction heuristic for the shared-path protection problem in wavelength division multiplexing networks with shared risk link group constraints and applies an evolutionary algorithm for optimizing the control parameters of the proposed heuristics. Thus, this algorithm is an example where the memetic algorithm, rather than trying to solve directly a given problem, employs an indirect approach whereby a heuristic method solves the problem while the memetic algorithm tries to improve the way in which the heuristic operates. Tang and Yao describe a memetic algorithm approach for a nonslicing and hard module very large scale integrated-circuit floorplanning problem. They tackle the issue of local search frequency [6] in the memetic algorithm using a bias search strategy. In their strategy, instead of improving all genetic individuals locally, only search points whose fitness values are above some threshold are exploited. It is shown that the memetic approach can quickly produce optimal or nearly optimal solutions for all the benchmark problems considered. Zhu et al. in “Wrapper-Filter Feature Selection Algorithm Using A Memetic Framework” describe a hybrid wrapper and filter feature selection algorithm for classification problems using a memetic framework. They incorporate a filter ranking method using a variety of information-theoretic and statistical measures as a local searcher within a genetic algorithm. They show that the proposed algorithm substantially improves classification accuracy while keeping the parsimony of the selected features. Several major issues on balancing local and genetic search in memetic algorithms are also considered. Tse et al.’s paper, “A Memetic Algorithm for MultipleDrug Cancer Chemotherapy Schedule Optimization,” describes a genetics-based memetic algorithm using iterative dynamic programming within the local search for the optimization of a multiple-drug cancer chemotherapy schedule. In this problem, the administration of a cocktail of drugs is formulated as an

3

optimal control problem by means of a set of dynamical equations. The goal is to minimize tumor size subject to a variety of constraints. This paper shows that the proposed algorithm is capable of solving a complex version of the multidrug cancer chemotherapy problem and that it is capable of dealing with multiple feasible solutions. Finally, in “Implementation of an Effective Hybrid GA for Large-Scale Traveling Salesman Problems” Nguyen et al. describe a parallel multipopulation steady-state memetic algorithm for solving large-scale traveling salesman problems (TSP) taking into consideration the tradeoff between computational time and solution quality. In their memetic algorithm, new variants of maximal preservative crossover, double-bridge move mutation, and the Lin–Kernighan local heuristic are used. It is shown that the proposed approach solves effectively and efficiently TSP benchmarks having up to 316 228 cities. Moreover, this paper provides a newfound best tour (as of June 2, 2003) for a 1 904 711-city TSP challenge. The papers collected in this special issue reflect the state of the art in the field of memetic algorithms. This special issue fulfils the original aim of showcasing the best research in the application of these algorithms to optimization and design problems. It is our hope that these papers will serve to inspire new ideas and provide enough challenging research issues for years to come.

ACKNOWLEDGMENT We would like to thank all the authors that submitted papers to this special issue, and we would also like to thank the stupendous efforts of the numerous referees that helped in reviewing and selecting these papers. We would like to give special thanks to D. Cook and L. Hall and their production team for their continuous support that made this special issue a great success.

YEW-SOON ONG, Guest Editor School of Computer Engineering Nanyang Technological University 639798 Singapore (e-mail: [email protected]) NATALIO KRASNOGOR, Guest Editor School of Computer Sciences and Information Technology Jubilee Campus University of Nottingham 1G8 1BB Nottingham, U.K. (e-mail: [email protected]) HISAO ISHIBUCHI, Guest Editor Department of Computer Science and Intelligent Systems Osaka Prefecture University Osaka 599-8531, Japan (e-mail: [email protected])

4

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 37, NO. 1, FEBRUARY 2007

R EFERENCES [1] Y. S. Ong, M. H. Lim, N. Zhu, and K. W. Wong, “Classification of adaptive memetic algorithms: A comparative study,” IEEE Trans. Syst., Man Cybern. B, Cybern., vol. 36, no. 1, pp. 141–152, Feb. 2006. [2] N. Krasnogor and J. Smith, “A tutorial for competent memetic algorithms: Model, taxonomy, and design issues,” IEEE Trans. Evol. Comput., vol. 9, no. 5, pp. 474–488, Oct. 2005. [3] H. Ishibuchi and T. Murata, “A multi-objective genetic local search algorithm and its application to flowshop scheduling,” IEEE Trans. Syst., Man Cybern. C, Appl. Rev., vol. 28, no. 3, pp. 392–403, Aug. 1998.

[4] J. D. Knowles and D. W. Corne, “Memetic algorithms for multiobjective optimization: Issues, methods and prospects,” in Recent Advances in Memetic Algorithms, W. E. Hart, N. Krasnogor, and J. E. Smith, Eds. New York: Springer-Verlag, 2005, pp. 313–352. [5] P. Merz, “Memetic algorithms for combinatorial optimization problems: Fitness landscapes and effective search strategies,” Ph.D. dissertation, Parallel Syst. Res. Group, Dept. Electr. Eng. Comput. Sci., Univ. Siegen, Siegen, Germany, 2000. [6] W. E. Hart, “Adaptive Global Optimization with Local Search,” Ph.D. dissertation, Univ. California, San Diego, CA, 1994.

Yew-Soon Ong (M’99) received the B.S. and M.S. degrees in electrical and electronics engineering from Nanyang Technology University, Jurong, Singapore, in 1998 and 1999, respectively, and the Ph.D. degree from the University of Southampton, Southampton, U.K., in 2002. He then joined the Computational Engineering and Design Centre, University of Southampton, U.K. He is currently an Assistant Professor with the School of Computer Engineering, Nanyang Technological University, Jurong, Singapore. His research interests lie in computational intelligence spanning: surrogate-assisted evolutionary algorithms, memetic algorithms, evolutionary robust design, evolutionary fuzzy systems, and grid computing. He is Guest Editor of Genetic Programming and Evolvable Machines Journal dedicated to Evolutionary Computation in Dynamic and Uncertain Environments. He has coedited a volume on advances in natural computation (Springer-Verlag) and editing an upcoming book dedicated to evolutionary computation in dynamic and uncertain environments in the Springer Series on Computational Intelligence. He is also member of the editorial board for the International Journal of Computational Intelligence, an Emergent Technologies technical committee member, and an Evolutionary Computation in Dynamic and Uncertain Environments technical committee member of the IEEE Computational Intelligence Society.

Natalio Krasnogor received the degree in systems analysis and informatics from the Universidad Nacional de La Plata, La Plata, Argentina, and the Ph.D. degree from the University of the West of England, Bristol, U.K. He did postdoctoral research in the University of Nottingham, Nottingham, U.K., where he became a Lecturer in 2002. He has established and cochaired the series of International Workshop on Memetic Algorithms (WOMA). In 2004, he was Guest Editor of a special issue of the Evolutionary Computation Journal dedicated to memetic algorithms. He is also a coeditor of the first book dedicated exclusively to recent advances in memetic algorithms published by SpringerVerlag. In 2004, he was Guest Editor of a special issue of the Journal of Fuzzy Sets and Systems dedicated to bioinformatics, and he is Guest Editor of an upcoming special issue on the theory and application of P-systems to systems biology for the Biosystems Journal. He is member of the editorial board of the International Journal of Computational Intelligence, and an Associate Editor of the Evolutionary Computation Journal. He is Principal Investigator or Coinvestigator in grants totaling in excess of 12 million pounds from the British research councils and the European Union. He currently supervises two postdoctoral fellows and six Ph.D. students. Dr. Krasnogor is a committee member of the Society for the Study of Artificial Intelligence and the Simulation of Behavior and is a member of the International Society for Genetic and Evolutionary Computation.

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 37, NO. 1, FEBRUARY 2007

5

Hisao Ishibuchi (M’93) received the B.S. and M.S. degrees in precision mechanics from Kyoto University, Kyoto, Japan, in 1985 and 1987, respectively, and the Ph.D. degree from Osaka Prefecture University, Osaka, Japan, in 1992. From 1987 to 2005, he was with Department of Industrial Engineering, Osaka Prefecture University. He is currently a Professor in Department of Computer Science and Intelligent Systems, Osaka Prefecture University. His research interests include evolutionary multiobjective optimization, multiobjective memetic algorithms, evolutionary game, genetic fuzzy systems, and fuzzy data mining. Dr. Ishibuchi received the Best Paper Award in the genetic algorithm track at the GECCO 2004 conference. He is an Associate Editor for the IEEE TRANSACTIONS ON FUZZY SYSTEMS, the IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B, and the IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE. He is also a Fuzzy Technical Committee Member of the IEEE Computational Intelligence Society.