Benchmarking ϵMAg-ES and BP-ϵMAg-ES on the bbob-constrained testbed
Michael Hellwig, Hans-Georg Beyer · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2022
The ϵMAg Evolution Strategie and its bi-population version (BP-ϵMAg-ES), which performed well on the CEC 2018 and CEC 2020 benchmark sets, are evaluated using the COCO bbob-constrained 2.6.2 benchmark release. Furthermore, the performance of Matlab's fmincon is considered as a baseline algorithm. Unlike the results of the CEC competitions, the outcomes obtained on the COCO 2.6.2 test set do not reveal any advantage of using a bi-population approach. The reasons for these observations needs still to be investigated in future research.