On the effect of environment-triggered population diversity compensation methods for memory enhanced UMDA

This paper focuses on the effect of population diversity to environment identification-based memory scheme (EI-MMS) which heuristically compensates population diversity through the storage and retrieving process of historic information. We introduced several diversity compensation measures and combined them with EI-MMS based univariate marginal distribution algorithm (UMDA) from two aspects. First, a basic diversity compensation measure was used to fight against the inherent diversity loss of UMDA. Second, two environment-triggered compensation measures were added in the sense of dynamic environment. Based on the experimental results on three dynamic test problems, the dynamics of population diversity of the corresponding EI-MMS based UMDAs were analyzed and several conclusions about how does the population diversity affect the performance of the algorithm in dynamic environments were drawn.