NeurIPS 2020

BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization

Meta Review

This paper presents BOTorch, an efficient Bayesian optimization library that enjoys several advantages over existing ones, such as a novel approach to optimize MC acquisition functions using fixed sample averages, a faster and easier implementation of second-order methods and look-ahead BO methods (e.g., Knowledge gradient), and general convergence results for sample-average approximation to acquisition functions via randomized quasi-MC. Overall, a good paper with strong results.