A Study of the Performance of $\text {Self-}{\star }$ Memetic Algorithms on Heterogeneous Ephemeral Environments

Abstract

We consider the deployment of island-based memetic algorithms (MAs) endowed with $\text {self-}{\star }$ properties on unstable computational environments composed of a collection of computing nodes whose availability fluctuates. In this context, these properties refer to the ability of the MA to work autonomously in order to optimize its performance and to react to the instability of computational resources. The main focus of this work is analyzing the performance of such MAs when the underlying computational substrate is not only volatile but also heterogeneous in terms of the computational power of each of its constituent nodes. We use for this purpose a simulated environment subject to different volatility rates, whose topology is modeled as scale-free networks and whose computing power is distributed among nodes following different distributions. We observe that in general computational homogeneity is preferable in scenarios with low instability; in case of high instability, MAs without self-scaling and self-healing perform better when the computational power follows a power law, but performance seems to be less sensitive to the distribution when these $\text {self-}{\star }$ properties are used.