Active Learning and Bayesian Optimization: A Unified Perspective to Learn with a Goal

• DOI: 10.1007/s11831-024-10064-z
• Publication Date: 2024-04-23T13:01:48Z
• Authors: Francesco Di Fiore; Michela Nardelli; Laura Mainini
• Keywords: Bayesian Optimization; Benchmark (surveying); Adaptive sampling

Abstract Science and Engineering applications are typically associated with expensive optimization problem to identify optimal design solutions states of the system interest. Bayesian active learning compute surrogate models through efficient adaptive sampling schemes assist accelerate this search task toward a given goal. Both those methodologies driven by specific infill/learning criteria which quantify utility respect set goal evaluating objective function for unknown combinations variables. While two fields have seen an exponential growth in popularity past decades, their dualism synergy received relatively little attention date. This paper discusses formalizes between as symbiotic common principles. In particular, we demonstrate unified perspective formalization analogy infill driving principles both goal-driven procedures. To support our original perspective, propose general classification techniques highlight similarities differences vast families sampling, learning, optimization. Accordingly, is demonstrated mapping criteria, formalized searches informed single information source multiple levels fidelity. addition, provide guidelines apply investigating performance different variety benchmark problems benefits limitations over mathematical properties that characterize real-world applications.