Distributed evolutionary algorithm topologies with adaptive migration schemes

Distributed evolutionary algorithms are of increasing interest and importance for three main reasons: (i) a well designed dEA can outperform a ´standard´ EA in terms of reliability, solution quality, and speed; (ii) they can (of course) be implemented on parallel hardware, and hence combine efficient utilization of parallel resources with very fast and reliable optimization; (iii) parallel hardware resources are increasingly common. A dEA operates as separate evolving populations with occasional interaction between them via ´migration´. A specific dEA is characterized by the topology and nature of these interactions. The performance of alternative topologies and migration mechanisms in this field remains under-explored. In this paper we continue an investigation of two simple, novel dEA topologies, comparing with the cube-based topology that underpins Alba et al´s GD-RCGA (a state of the art dEA). The focus in this paper is on testing a novel adaptive migration scheme, in which the frequency of migration events adapts dynamically in response to the current balance between exploration and exploration. We also focus on high dimensional versions of a selection of hard function optimization problems. We find that the adaptive migration scheme is promising, and that overall results marginally favour a simple three-level tree based topology and adaptive migration with a longer window, especially as dimensionality increases.