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

Abstract

Both Bayesian optimization and active learning realize an adaptive sampling scheme to achieve a specific learning goal. However, while the two fields have seen an exponential growth in popularity in the past decade, their dualism has received relatively little attention. In this paper, we argue for an original unified perspective of Bayesian optimization and active learning based on the synergy between the principles driving the sampling policies. This symbiotic relationship is demonstrated through the substantial analogy between the infill criteria of Bayesian optimization and the learning criteria in active learning, and is formalized for the case of single information source and when multiple sources at different levels of fidelity are available. We further investigate the capabilities of each infill criteria both individually and in combination on a variety of analytical benchmark problems, to highlight benefits and limitations over mathematical properties that characterize realworld applications.

Self-Driving Laboratories for Chemistry and Materials Science

“…One of the most common strategies today for SDL experimental planning is Bayesian optimization (BO), which aims to maximize or minimize some black-box function, such as a measurable chemical property, as a function of controllable experimental parameters. 288,289 To do this, the surrogate model with some prior distribution is fit, or trained, on the available data, and the predicted posterior distribution is used to create an acquisition function. The acquisition function contains information about the prediction and the uncertainty of the prediction based on the posterior distribution, and can be used to control the exploitative or explorative nature of the subsequent experiments.…”

Tackling Data Scarcity Challenge through Active Learning in Materials Processing with Electrospray

“…Under the umbrella of adaptive learning, AL and BO are both goal-driven learning strategies with different focuses. [29] BO, or more broadly surrogate-based optimization (SBO), placed emphasis on the identification of optimum candidates. [30] SBO algorithms seek balance of exploration and exploitation to avoid being trapped in local minimum.…”