Modeling Coevolutionary Genetic Algorithms on Two-Bit Landscapes: Random Partnering

Ming Chang¹, Kazuhiro Ohkura², Kanji Ueda³, and Masaharu Sugiyama¹

¹Gifu Prefecture Research Institute of Manufacturing Information Technology, 19-179-4 Sue, Kakamigahara city, Gifu 509-0108, Japan
chang@gifu-irtc.go.jp
sugi@gifu-irtc.go.jp

²Faculty of Engineering, Kobe University, 1-1 Rokkoda-cho, Nada-ku, Kobe 657-8501, Japan.
ohkura@mech.kobe-u.ac.jp

³Research Into Artifacts Center for Engineering, The University of Tokyo 4-6-1, Komaba, Meguro, Tokyo 153-8904 Japan.
ueda@race.u-tokyo.ac.jp

Abstract. A model of coevolutioinary genetic algorithms (COGA) consisting of two populations coevolving on two-bit landscapes is investigated in terms of the effects of random partnering strategy, different population updating schemes, and changes in mutation rate and evolution rate. The analytical and numerical approaches showed that even in such a simple model, the dynamics can change dramatically with different evolutionary scenarios in such an extent that deserves our attention from the point of view of algorithm design.

LNCS 3102, p. 513 ff.

Full article in PDF

lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004