An Asynchronous Adaptive Multi-population Model for Distributed Differential Evolution

In this paper a general-purpose asynchronous adaptive multi-population model for distributed Differential Evolution (AsAMP-dDE) algorithm is proposed. The distributed algorithm, following the stepping-stone model, is characterized by an asynchronous mechanism for the migration and for a multi-population recombination employed to exchange information. The adaptive procedure is based on two steps. Firstly a local performance measure related to the average fitness improvement for each subpopulation is computed. Secondly, a specific updating scheme based on these measures takes place to randomly update the control parameter values. The asynchronous migration mechanism and the adaptive procedure allow reducing the number of control parameters to be set in the distributed model. AsAMP-dDE has been tested on the benchmarks of the CEC2016 real parameter single objective competition without adopting any specific mechanism opportunely tailored for solving such test problems. The results show that this algorithm allows obtaining good performance in most of the investigated benchmarks.