3207 大域的最適化のための複数最適化手法の動的制御法(OS4 最適化計算法(I),未来社会を支えるものづくりとひとづくり(設計・システムから))

概要

論文の詳細を見る
When we apply the optimization algorithms to target problems, it is necessary to select the appropriate algorithms according to the characteristic of the problems. Moreover, the control parameters of the algorithms significantly affect the search ability. Therefore, it is also necessary to determine the appropriate parameters. However, the strategy for solving these difficulties depends on the knowledge and experience of users. Therefore, it is not easy for ordinary designers. In this research, we propose the strategy for dynamically controlling plural optimization methods. There are two features of the proposed strategy ; 1) the maximum time to execute the optimization algorithms is defined, and the purpose of the strategy is to obtain the better solutions for the limited time, and 2) the landscape of the search space can be roughly grasped after the optimization. We show that the proposed strategy is a effective tool for the design problems.
2006-11-14