Multi-population univariate marginal distribution algorithm for dynamic optimization problems

An improved multi-population univariate marginal distribution algorithm (MUMDA) is proposed to solve dynamic optimization problems. The search space is divided into several parts by using several probability modals which correspond to several populations. Meanwhile, the algorithm explores and exploits in different regions and the best solutions are migrated. The objective is to enlarge the search space, increase the population diversity and adapt to the change of the environments rapidly. Moreover, the convergence of UMDA is proved, which is used to analyze the validity of the proposed algorithm. Finally, an experimental study is carried out to compare the performance of several UMDA. The experimental results show that the MUMDA is effective and can adopt the dynamic environments rapidly.