Distributed differential evolution with explorative-exploitative population families

This paper proposes a novel distributed differential evolution algorithm, namely Distributed Differential Evolution with Explorative-Exploitative Population Families (DDE-EEPF). In DDE-EEPF the sub-populations are grouped into two families. Sub-populations belonging to the first family have constant population size, are arranged according to a ring topology and employ a migration mechanism acting on the individuals with the best performance. This first family of sub-populations has the role of exploring the decision space and constituting an external evolutionary framework. The second family is composed of sub-populations with a dynamic population size: the size is progressively reduced. The sub-populations belonging to the second family are highly exploitative and are supposed to quickly detect solutions with a high performance. The solutions generated by the second family then migrate to the first family. In order to verify its viability and effectiveness, the DDE-EEPF has been run on a set of various test problems and compared to four distributed differential evolution algorithms. Numerical results show that the proposed algorithm is efficient for most of the analyzed problems, and outperforms, on average, all the other algorithms considered in this study.