Distributed mixed variant differential evolution algorithms for unconstrained global optimization

This paper proposes a novel distributed differential evolution algorithm called Distributed Mixed Variant Differential Evolution (dmvDE). To alleviate the time consuming trial-and-error selection of appropriate Differential Evolution (DE) variant to solve a given optimization problem, dmvDE proposes to mix effective DE variants with diverse characteristics in a distributed framework. The novelty of dmvDEs lies in mixing different DE variants in an island based distributed framework. The 19 dmvDE algorithms, discussed in this paper, constitute various proportions and combinations of four DE variants (DE/rand/1/bin, DE/rand/2/bin, DE/best/2/bin and DE/rand-to-best/1/bin) as subpopulations with each variant evolving independently but also exchanging information amongst others to co-operatively enhance the efficacy of the distributed DE as a whole. The dmvDE algorithms have been run on a set of test problems and compared to the distributed versions of the constituent DE variants. Simulation results show that dmvDEs display a consistent overall improvement in performance than that of distributed DEs. The best of dmvDE algorithms has also been benchmarked against five distributed differential evolution algorithms. Simulation results reiterate the superior performance of the mixing of the DE variants in a distributed frame work. The best of dmvDE algorithms outperforms, on average, all five algorithms considered.