On the impact of active covariance matrix adaptation in the CMA-ES with mirrored mutations and small initial population size on the noiseless BBOB testbed
Dimo Brockhoff, Anne Auger, Nikolaus Hansen · 2012
Mirrored mutations as well as active covariance matrix adaptation are two techniques that have been introduced into the well-known CMA-ES algorithm for numerical optimization. Here, we investigate the impact of active covariance matrix adaptation in the IPOP-CMA-ES with mirrored mutation and a small initial population size. Active covariance matrix adaptation improves the performance on 8 of the 24 benchmark functions of the noiseless BBOB test bed. The effect is the largest on the ill-conditioned functions with the largest improvement on the discus function where the expected runtime is more than halved. On the other hand, no statistically significant adverse effects can be observed.