Source: Deep Learning on Medium

There are interesting links between Bayesian optimization and reinforcement learning. Specifically, Bayesian optimization is a sequential decision problem where the decisions (choices of x to evaluate) do not affect the state of the system (the actual function f). Such state-less sequential decision problems fall under the rubric of multi-arm bandits72, a subclass of reinforcement-learning problems. More broadly, important recent work takes a Bayesian approach to learning to control uncertain systems73 (for a review see ref. 74).