Multi-Objective Algorithm Configuration on Multi-Modal Multi-Objective Optimization Problems

This thesis successfully investigated interesting aspects of multi-objective (MO) automated algorithm configuration (AAC) for multi-modal multi-objective optimization problems (MOP) and gained interesting insights into a so far rather unexplored field at the forefront of current research. In this context, various evolutionary multi-objective algorithms (EMOA) were configured to simultaneously generate diverse solutions in the decision space and foster convergence towards the Pareto front. These configurations were performed using the model-based AAC algorithms (MO-)SMAC which demonstrated their high performance and potential for multi-objective configuration tasks. As test instances, a set of, also multi-modal, multi-objective optimization instances of different benchmark function collections were utilized.