Deep Learning-Assisted Evolutionary Meta-Strategy Search Considering the Balance of Versatility and Efficiency

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K11968
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61040:Soft computing-related
• Research Institution: Muroran Institute of Technology
• Principal Investigator: Watanabe Shinya 室蘭工業大学, 大学院工学研究科, 准教授 (30388136)
• Co-Investigator(Kenkyū-buntansha): 榊原 一紀 富山県立大学, 工学部, 准教授 (30388110)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 進化計算 / 進化型多目的最適化 / 主成分分析 / ベイズ最適化

Outline of Final Research Achievements

In this study, we investigated meta-strategies aimed at achieving versatility and efficiency in evolutionary computation for both single and multiple objective problems. Our research involved the development of two distinct approaches. Firstly, we proposed a meta-strategy that dynamically utilizes principal component analysis based on the specific situation. Secondly, we introduced a meta-strategy that dynamically adjusts the balance between local and global exploration in Bayesian optimization. The former approach was designed to address general multi-objective optimization problems, while the latter was tailored for single-objective optimization problems in environments with limited computational resources.
To evaluate the effectiveness of these approaches, we conducted experiments by applying them to representative test problems. The results obtained from these validation experiments demonstrated the efficacy of our proposed meta-strategies across a wide range of problem domains.

Free Research Field

情報工学

Academic Significance and Societal Importance of the Research Achievements

本研究では，単目的および多目的問題のための汎用性と効率性を両立した進化計算のためのメタ戦略アプローチを開発した．本成果は，進化計算および最適化の適用が難しかった多くの実問題において，有効な解決策となると考えている．高い汎用性を持つ進化計算はこれまで幅広い問題に応用されてきたものの，シミュレーション負荷が高くあまり試行錯誤が許容されない実問題への適用は難しかった．本研究成果は，その解決につながるものであり，その社会的意義は大きいと考えている．