Ensemble centralities based adaptive Artificial Bee algorithm

An adaptive Artificial Bee Colony algorithm based on centralities is presented in this paper. As complex networks are generated in evolutionary algorithms during iterations, it becomes possible to obtain meaningful information regarding population dynamics during evaluations. The three centralities of Degree, Closeness and Betweenness are used for adaptive population control of the algorithm, where population interaction is measured and least performing solutions are replaced. Two adaptive variants of the algorithm are presented, one based on a single population and the other on an ensemble population approach. The experimentation is conducted on various standard test functions, showing that the adaptive approaches offer an improvement upon the canonical algorithm.