Masaharu Munetomo and Hiroshi Someya

The Journal of The Institute of Electrical Engineers of Japan, Vol. 132, No. 4, pp.204-207, Special Issue on Practical Applications of Evolutionary Computation, joi: JST.JSTAGE/ieejjournal/132.204  (in Japanese), Apr (2012)

Keywords: Evolutionary Computation, Genetic Algorithm, Adaptive Monte Carlo Method, Guideline on Applying Evolutionary Algorithms, Advanced Techniques

Full Text: publisher-created version

Advertisements

Hiroshi Someya, Hisashi Handa and Seiichi Koakutsu

IEEJ Transactions on Electronics, Information and Systems, Vol. 132, No. 1, pp.2-5, Special Issue Review, doi: 10.1541/ieejeiss.132.2  (in Japanese with English abstract), Jan (2012)

Abstract: This article presents a review of recent advances in stochastic optimization algorithms. Novel algorithms achieving highly adaptive and efficient searches, theoretical analyses to deepen our understanding of search behavior, successful implementation on parallel computers, attempts to build benchmark suites for industrial use, and techniques applied to real-world problems are included. A list of resources is provided.

Keywords: Evolutionary Computation, Swarm Intelligence, Differential Evolution, Tabu Search, Real-world Application

Full Text: publisher-created version

Hiroshi Someya

Soft Computing, Vol. 16, No. 1, pp. 23-45, doi: 10.1007/s00500-011-0732-1, Jan 2012

Abstract: Studies on parameter tuning in evolutionary algorithms are essential for achieving efficient adaptive searches. This paper discusses parameter tuning in real-valued crossover operators theoretically. The theoretical analysis is devoted to improving robustness of real-coded genetic algorithms (RCGAs) for finding optima near the boundaries of bounded search spaces, which can be found in most real-world applications. The proposed technique for crossover-parameter tuning is expressed mathematically, and thus enables us to control the dispersion of child distribution quantitatively. The universal applicability and effect have been confirmed theoretically and verified empirically with five crossover operators. Statistical properties of several practical RCGAs are also investigated numerically. Performance comparison with various parameter values has been conducted on test functions with the optima placed not only at the center but also in a corner of the search space. Although the parameter-tuning technique is fairly simple, the experimental results have shown the great effectiveness.

Keywords: real-parameter evolutionary algorithms, parameter setting, robustness, bounded search space, numerical optimization

Full Text: publisher-created version

Hiroshi Someya

IEEE Transactions on Evolutionary Computation, Vol. 15, No. 2, pp. 248-266, doi: 10.1109/TEVC.2010.2083668, April 2011.

Abstract: This paper investigates the evolutionary dynamics of steady-state real-valued evolutionary algorithms (RVEAs) with more-than-one-element replacement theoretically, whereas most theoretical studies of RVEAs have considered single- or all-element replacement. The subject RVEAs are of interest because they appear in various fashions, such as real-coded genetic algorithms (RCGAs) and island RVEAs. The analysis is conducted to deepen the understanding of how RVEA components and their parameters influence the phenotypic diversity in the parental pool. Firstly, the diversity evolution is modeled mathematically and then a constraint of diversity control is derived from this model. The control method is demonstrated and the accuracy of the theoretical predictions is evaluated through experiments. The shortest convergence time is estimated. The analysis requires few assumptions about either the variation operators or selection schemes, and therefore is applicable to various RVEAs. As such an application in RCGAs, the influence on the diversity evolution of offspring-population size, parental-pool size, crossover-operator parameter and selection-pressure parameters of two selection mechanisms is quantified. The computational efficiency, search stability and selection-pressure controllability are then evaluated. The analysis results are discussed from a practical point of view in parameter settings for preventing premature convergence.

Full Text: publisher-created version

Takashi Onoda and Hiroshi Someya

IEEJ Transactions on Electronics, Information and Systems, Vol.130, No.1, pp.2–5, Special Issue Review, doi:10.1541/ieejeiss.130.2 (in Japanese with English abstract) (January 2010).

Abstract: This article surveys the history of the field of Neural Network research and presents a review of several techniques developed in the field. Attempts at statistical analysis of search dynamics of the optimization methods in Soft Computing and recent advances on implementation in parallel computers are briefly introduced.

Full Text: publisher-created version

Hiroshi Someya, Kensaku Sakamoto, and Masayuki Yamamura

In Proceedings of ACM Genetic and Evolutionary Computation Conference: GECCO-2009, pp.233-240, doi:10.1145/1569901.1569934, Montreal, Canada (July 2009).

Abstract: Protein engineering, developing novel proteins with a desired activity, has become increasingly important in many fields. This paper presents two studies in protein engineering: (i) a biological implementation of a genetic algorithm, with an observed in vitro evolution, and (ii) its preliminary computer simulation using a prototypical probabilistic model based on a random walk. The steady evolution of the fitness distribution of the mutant proteins that appeared in the biological experiments has provided some convincing evidence about the search behavior and the fitness landscape. The computer simulation and the simple probabilistic model have indicated their future potential for providing a practical alternative to the time-consuming manual operations in the biological experiments. Successful experimental results in the two studies have raised expectations of their further development and mutually beneficial interactions.

Full Text: publisher-created version

Hiroshi Someya

Lecture Notes in Computer Science 5361 Simulated Evolution and Learning (Proceedings of The Seventh International Conference on Simulated Evolution And Learning: SEAL2008), pp.269–278, doi:10.1007/978-3-540-89694-4_28, Melbourne, Australia (December 2008).

Abstract: Parameters of real-valued crossover operators have been often tuned under a constraint for preserving statistics of infinite parental population. For applications in actual scenes, in a previous study, an alternative constraint, called unbiased constraint, considering finiteness of the population has been derived. To clarify the wide applicability of the unbiased constraint, this paper presents two additional studies: (1) applying it to various crossover operators in higher dimensional search space, and (2) generalization of it for preserving statistics of overall population. Appropriateness of the parameter setting based on the unbiased constraint has been supported in discussion on robust search.

Full Text: publisher-created version