Parallel evolutionary algorithms

Abstract

Evolutionary algorithms (EAs) have given rise to many parallel variants, fuelled by the rapidly increasing number of CPU cores and the ready availability of computation power through GPUs and cloud computing. A very popular approach is to parallelize evolution in island models, or coarse-grained EAs, by evolving different populations on different processors. These populations run independently most of the time, but they periodically communicate genetic information to coordinate search. Many applications have shown that island models can speed up computation significantly, and that parallel populations can further increase solution diversity. The aim of this book chapter is to give a gentle introduction into the design and analysis of parallel evolutionary algorithms, in order to understand how parallel EAs work, and to explain when and how speedups over sequential EAs can be obtained. Understanding how parallel EAs work is a challenging goal as they represent interacting stochastic processes, whose dynamics are determined by several parameters and design choices. This chapter uses a theory-guided perspective to explain how key parameters affect performance, based on recent advances on the theory of parallel EAs. The presented results give insight into the fundamental working principles of parallel EAs, assess the impact of parameters and design choices on performance, and contribute to an informed design of effective parallel EAs.