Overview of Surrogate-model Versions of Covariance Matrix Adaptation Evolution Strategy

Abstract

Evaluation of real-world black-box objective functions is in many optimization problems very time-consuming or expensive. Therefore, surrogate regression models, used instead of the expensive objective function and in that way decreasing the number of its evaluations, have received a lot of attention. Here, we briefly survey surrogate-assisted versions of the state-of-the-art algorithm for continuous black-box optimization - the CMA-ES (Covariance Matrix Adaptation Evolution Strategy). We compare five of them, together with the original CMA-ES, on the noiseless benchmarks of the Comparing-Continuous-Optimisers platform in the expensive scenario, where only a small budget of evaluations is available.