An operation to promote diversity in evolutionary algorithms in a dynamic hybrid island model

ABSTRACTCurrently, there is a considerable variety of Evolutionary Algorithms (EAs) and due to their performances some of them become more popular. EAs can be implemented in different ways, such as the Island Model (IM). However, despite the good performance of some EAs and the possibilities of varying their implementations, they can converge to a local optimum mainly because of the loss of diversity in the population. This work proposes an operation for a dynamic hybrid IM (D-IM), aiming to promote diversity to the population if it is converging to a certain portion of the search space. Thus, the D-IM reacts to the possible local convergence of its population, in addition to adjust the topology according to the EAs in the islands. The results demonstrated that the proposed operation can improve the efficiency of the D-IM search process and be competitive for solving bounded constrained optimization problems.