Development of Canonical Particle Swarm Optimization based on Dynamical System Theory

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K06077
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Communication/Network engineering
• Research Institution: Nippon Institute of Technology
• Principal Investigator: Jin'no Kenya 日本工業大学, 工学部, 教授 (50286762)
• Co-Investigator(Kenkyū-buntansha): 進藤 卓也 日本工業大学, 工学部, 助教 (90749110) ; 木村 貴幸 日本工業大学, 工学部, 准教授 (80579607)
• Research Collaborator: Hariya Yosuke
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 群知能 / 決定論システム / 非線形力学系 / 連続値最適化 / 機械学習 / PWMインバータ / 不変性 / 悪スケール性

Outline of Final Research Achievements

We developed a nonlinear mapping optimization method with invariance based on analysis result of invariance of solution search performance due to variable dependency and ill-scale. Although the proposed algorithm is a deterministic system, local search capability improves and it has invariance. Based on the preliminary simulation results, we confirmed that the nonlinear mapping optimization method has higher search performance than the particle swarm optimization. When the proposed algorithm applies to machine learning, we confirmed that efficient learning result is obtained. The PWM inverter that is designed by using the proposed optimization method exhibits high power conversion efficiency.

Free Research Field

非線形工学