Abstract
In this study, we propose a new multi-point type quasi-chaotic optimization method that can be implemented without a vital parameter tuning. The multi-point type quasi-chaotic optimization method (M-Q-COM), which has been proposed recently, is a global optimization method to solve unconstrained optimization problems in which the simultaneous perturbation gradient approximation is introduced into a multi-point type chaotic optimization method. The M-Q-COM can be applied to a class of problems whose objective function values only can be computed. It is very effective to solve unconstrained multipeaked optimization problems with 100 variables. However, in the M-Q-COM, a parameter called initial sampling parameter must be tuned to generate quasi-chaotic search trajectory that is used to implement global search. In this study, we introduce two approaches to deal the issue. In the first approach, a descent sign vector extracted from an approximated gradient vector is used as a search direction of a search point. Then, its moving distance is determined so that its search process conforms to the search process of the quasi-chaotic optimization method obligatorily. The other approach is an initial sampling parameter estimation. We confirm effectiveness of the proposed methods through applications to the aforementioned benchmark problems.
