設計変数間の依存関係を反映したハイブリッドメタ戦略による多峰性問題の整数最適化  [in Japanese] The Hybrid Meta-heuristics by Reflecting Recognition of Dependence Relation among Design Variables for The Integer Optimization of Multi-peak Problems  [in Japanese]

In this paper, Authors proposed a hybrid meta-heuristic method for the integer optimization of multi-peak problems to support the engineering field. For the search ability of the optimal process, the proposed method did not define a traditional combination, i.e., combining global search ability and local search ability, but newly defined the combination of multi search ability (MSA), local search ability (LSA) and neighborhood optimal search ability (NOSA). For NOSA, this proposed method introduced direct search process which self-changes design variables with reflecting recognition of dependence relation among design variables automatically. The proposed method was applied to some benchmark functions to estimate its performance, and was confirmed that it improved amount of computational cost and convergence up to the optimal solution.