Estimation of distribution algorithm (EDA) is one of famous evolutionary algorithms (EAs) and has successfully solved various problems. However, EDA usually needs large population; specifically, the population size increase exponentially with the degree of interactions among sub-functions. We use a new statistical test approach, called FastANOVA, to find the linkage relation with small sample size. In this study, we conduct experiments to examine the linkage accuracy using trap-k problem and examine the usability by CEC2014 benchmark and uses NK landscape problem to examine the advantage of using FastANOVA in EDA. The experimental results show the FastANOVA improves the performance of EDA using small population.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。