Abstract
The population structure of an evolutionary algorithm influences the dissemination and mixing of advantageous alleles, and therefore affects search performance. Much recent attention has focused on the analysis of complex population structures, characterized by heterogeneous connectivity distributions, non-trivial clustering properties, and degree-degree correlations. Here, we synthesize the results of these recent studies, discuss their limitations, and highlight several open questions regarding (i) unsolved theoretical issues and (ii) the practical utility of complex population structures for evolutionary search. In addition, we will discuss an alternative complex population structure that is known to significantly influence dynamical processes, but has yet to be explored for evolutionary optimization. We then shift our attention toward dynamic population structures, which have received markedly less attention than their static counterparts. We will discuss the strengths and limitations of extant techniques and present open theoretical and experimental questions and directions for future research. In particular, we will focus on the prospects of "active linking," wherein edges are dynamically rewired according to the genotypic or phenotypic properties of individuals, or according to the success of prior inter-individual interactions.