Automatic Configuration of Preference-based Evolutionary Multi-objective Optimization Algorithms

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K17824
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61040:Soft computing-related
• Research Institution: Yokohama National University
• Principal Investigator: Tanabe Ryoji 横浜国立大学, 大学院環境情報研究院, 助教 (80780923)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 進化計算 / 進化型多目的最適化 / 選好に基づく最適化 / 自動アルゴリズム生成

Outline of Final Research Achievements

Multi-objective optimization aims to simultaneously minimize multiple objective functions that conflict with each other. Multi-objective optimization problems can be found in a wide range of engineering applications. Although a preference-based multi-objective evolutionary algorithm is a useful approach to finding a set of non-dominated solutions preferred by a decision-maker, its performance strongly depends on the algorithmic configuration. In addition, it is difficult to optimize the algorithm configuration by means of hand-tuning. To address this issue, this work studied a framework for automated generation of preference-based multi-objective evolutionary algorithms. Secondarily, this work also addressed benchmarking issues in preference-based multi-objective evolutionary algorithms.

Free Research Field

進化計算

Academic Significance and Societal Importance of the Research Achievements

本研究で開発した選好に基づく多目的進化型アルゴリズムを自動生成する枠組みを利用すれば, 非専門家であるユーザでも手軽に最適化が可能となった. また, 自動アルゴリズム構成は研究者がこれまで思いつかなかったような構成を生成する場合が多い. そのため, 高性能かつ新規性のある多目的進化型アルゴリズムが自動生成されることが期待でき, 本研究分野に新しい視点をもたらすことが期待される.