Self-organization evolutionary algorithm for dynamic optimization problems

Dynamic optimization problems require an algorithm to continuously track a changing optimal solution over time. In this paper, a self-organization univariate marginal distribution algorithm (SOUMDA) is proposed to solve dynamic optimization problems. The self-organization scheme is used to increase the diversity of population in a guided fashion. The self-organization scheme consists of two parts. One part uses the history information of optimal solution to predict the change direction, and the other part uses the current information to diverge from the local optimum. Finally, an experiment on dynamic sphere function is carried out to compare the performances of several UMDAs. The experimental results show that the SOUMDA is effective and can adapt to the dynamic environments rapidly.