Optimization Problems with Diverse Structures Based on Quality-Diversity
| Reference No. | 2025a045 |
|---|---|
| Type/Category | Grant for Project Research-Short-term Joint Research |
| Title of Research Project | Optimization Problems with Diverse Structures Based on Quality-Diversity |
| Principal Investigator | Shoichiro Tanaka（The University of Fukuchiyama・Assistant professo） |
| Research Period |
August 18,2025. -
August 22,2025. |
| Keyword(s) of Research Fields | Reeb graph, Manifold, Quality-Diversity optimization, Machine learning |
| Abstract for Research Report | Optimization problems in the real world often do not have explicit objective functions because candidate solutions are evaluated through experiments or simulations. Such optimization problems are called black-box optimization problems. The structure of black-box optimization problems—such as the number of optima, their distribution, and topology—is a priori unknown. Since the nature of the problem structure strongly affects the performance of optimization algorithms, selecting an appropriate optimization algorithm for problems with unknown structure is difficult. Landscape analysis is a method for understanding and analyzing the structure of black-box optimization problems. So far, landscape analysis has visualized and analyzed the structures of various optimization problems. However, while landscape analysis reveals the structure of a given problem, it cannot design optimization problems with specific structures. To investigate the performance of optimization algorithms, we want to design optimization problems with diverse structures. Therefore, this project aims to construct a set of optimization benchmark problems with expected structures. Using a new paradigm of evolutionary computation, we will explore optimization problems that achieve both quality as optimization problems and diversity of structures. |
| Organizing Committee Members (Workshop) Participants (Short-term Joint Usage) |
Shoichiro Tanaka（The University of Fukuchiyama・Assistant professor） Daisuke Sakurai（Fujitsu Limited・Research director） |