GPU-based automatic configuration of differential evolution: A case study

Abstract

The performance of an evolutionary algorithm strongly depends on the choice of the parameters which regulate its behavior. In this paper, two evolutionary algorithms (Particle Swarm Optimization and Differential Evolution) are used to find the optimal configuration of parameters for Differential Evolution. We tested our approach on four benchmark functions, and the comparison with an exhaustive search demonstrated its effectiveness. Then, the same method was used to tune the parameters of Differential Evolution in solving a real-world problem: the automatic localization of the hippocampus in histological brain images. The results obtained consistently outperformed the ones achieved using manually-tuned parameters. Thanks to a GPU-based implementation, our tuner is up to 8 times faster than the corresponding sequential version. © 2013 Springer-Verlag.