“…A common drawback of this family of algorithms is that reduction to single-objective optimization can potentially lead to more exploitation behavior resulting in sub-optimal solutions. PAL (Zuluaga, Sergent, Krause, & Püschel, 2013), PESMO (Hernández-Lobato et al, 2016), and the concurrent works USeMO (Belakaria, Deshwal, Jayakodi, & Doppa, 2020b) and MESMO (Belakaria et al, 2019) are principled algorithms based on information theory. PAL tries to classify the input points based on the learned models into three categories: Pareto optimal, non-Pareto optimal, and uncertain.…”