Analysis and Switching of Normalization in the Decomposition-based Constraint Handling Technique for Constrained Optimization

カテゴリ: 部門大会

論文No: SS2-7

グループ名: 【C】2023年電気学会電子・情報・システム部門大会

発行日: 2023/08/23

タイトル(英語): Analysis and Switching of Normalization in the Decomposition-based Constraint Handling Technique for Constrained Optimization

著者名: Yasuda Yusuke（Tokyo Metropolitan University）,Kojima Hidetoki（JGC Japan）,Kumagai Wataru（Yokogawa Electric）,Tamura Kenichi（Tokyo Metropolitan University）,Yasuda Keiichiro（Tokyo Metropolitan University）

著者名(英語): Yusuke Yasuda (Tokyo Metropolitan University),Hidetoki Kojima (JGC Japan),Wataru Kumagai (Yokogawa Electric),Kenichi Tamura (Tokyo Metropolitan University),Keiichiro Yasuda (Tokyo Metropolitan University)

キーワード: Constrained Optimization|Metaheuristics|Constraint Handling|MOEA/D|Normalization|Constrained Optimization|Metaheuristics|Constraint Handling|MOEA/D|Normalization

要約(日本語): Our recent work extended Multiobjective Evolutionary Algorithm based on Decomposition (MOEA/D) to handle constrained single-objective optimization and proposed MOEA/D with adaptive weight adjustment. This paper indicates that the normalization used in our recent work leads to poor search performance and proposes a switching normalization method that updates/stops the reference points of normalization depending on the search situation. Numerical experiments show that the proposed method enhances the robustness and performance of the search.