國 立 陽 明 交 通 大 學 演 講

講    題：Particle Swarm Optimization with Pseudo-random number Assignment Strategy

主 講 人：遲銘璋 (國立中興大學企業管理學系 副教授)

主 持 人：董弘平教授

主辦單位：陽明交通大學工業工程與管理系

時間：2022年11月28日(星期一) 13:20~15:20

地    點：綜合大樓AB103館

演講摘要：Particle swarm optimization (PSO) is a popular metaheuristic algorithm, and it belongs to a stochastic population-based optimization method. The random search scheme in PSO intended for simulating social behavior is primarily driven by a couple of random streams utilized in the particle velocity equation, i.e., individual and social randomness influences. The pseudo random assignment (PRN) strategy is an advanced simulation technique based on manipulating the random stream sets that drive the stochastic metaheuristic method or simulation model. In this talk, we discuss the influences of introducing three different PRN strategies into PSO algorithm. Our algebraic analysis is presented using stochastic process theory to understand the stability regions and conditions. The PSO with different PRN strategies are tested and evaluated on eight unimodal and multimodal benchmark functions. Six different definitions of swarm diversity are used to measure the rate of change of swarm diversity. Our analytic results indicate that the different PRN strategy has no effect on the expectation, but it would reduce or increase the variance of particle dynamics. The empirical results reveal that PSO with common random stream is superior to other PRN strategies in terms of the mean best solution found at the end of simulation.

演講性質：學術研究專題

歡迎聽講