On the potential of automated algorithm configuration on multi-modal multi-objective optimization problems
Jeroen Rook, Heike Trautmann, Jakob Bossek, Christian Grimme · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2022
Hardness of Multi-Objective (MO) continuous optimization problems results from an interplay of various problem characteristics, e. g. the degree of multi-modality. We present a benchmark study of classical and diversity focused optimizers on multi-modal MO problems based on automated algorithm configuration. We show the large effect of the latter and investigate the trade-off between convergence in objective space and diversity in decision space.