Testing the impact of parameter tuning on a variant of IPOP-CMA-ES with a bounded maximum population size on the noiseless BBOB testbed
Tianjun Liao, Thomas Stützle · 2013
In this paper, we experimentally explore the influence tuned parameter settings have on an IPOP-CMA-ES variant that uses a maximum bound on the population size. We followed our earlier work, where we exposed seven parameters that control parameters of IPOP-CMA-ES, and tune them by applying irace, an automatic algorithm configuration tool. A comparison of the tuned to the default settings on the BBOB benchmark shows that for difficult problems such as multi-modal functions with weak global structure, the tuned parameter settings can result in significant improvements over the default settings.