[4L3-J-8] Soft computing

[4L3-J-8-04] A study on search method considering search difficulty of objective functions

〇Kento Niwa¹, Tomohiro Yoshikawa¹ (1. Nagoya University)

Keywords:multi-objective optimization problem, search difficulty, search resource allocation

In real-world multi-objective problems, the characteristics of objective functions and their search difficulties are different one another. Therefore, when search resources are evenly allocated like MOEA/D, the objective functions with low search difficulty are preferentially optimized while those with high search difficulty are not optimized. In real-world problems, it is better to search equally for each objective function. In this paper, we propose a method to allocate search resources based on their search difficulties. We performed quantitative evaluation on optimization balance between objective functions on 2 and 4 objective optimization problem. The results of experiments showed the proposed method could perform more balanced search than conventional methods.