Effect of communication topologies on hybrid evolutionary algorithms

Multi-population bio-inspired algorithms present attractive potential for hybridization because of the relatively low degree of coupling they require between groups. In this work, we present a multiple swarm particle swarm optimization (MPSO) algorithm that has been modified to incorporate populations from a genetic algorithm. We investigate the ways in which the performance of this hybrid algorithm is influenced by the topological strategy that is used to direct communication between populations. The results suggest that in addition to the topological layout, the placement of different types of swarms may indirectly affect the resulting solution quality. The hybrid algorithm with varying communication topologies is implemented on a GPU architecture.