Guiding the next experiment: Bayesian Global Optimization versus Reinforcement Learning

Back

Guiding the next experiment: Bayesian Global Optimization versus Reinforcement Learning

Turab Lookman

AiMaterials Research LLC, USA

With the development of self-driving laboratories, Bayesian Global Optimization (BGO) has been the method of choice in many recent studies since its use in accelerated materials science in 2015. I will show how Reinforcement Learning can be applied to accelerate discovery with application to solid phase change alloys. Moreover, I will present validation results on synthetic optimizing functions that indicate the relative merits of the approaches as a function of the number of descriptors, number of experiments, batch size, and how the methods seek high value regions of the objective/property as the experimental iterations proceed.