Turab Lookman1 2 3*
1 AiMaterials Research LLC, Santa Fe, 87501, USA；
2 State Key Laboratory for Mechanical Behavior of Materials, Xi’an Jiaotong University, Xi’an, 710049, China；
3 Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing, Beijing, 100083, China；
Abstract：Sequential Design lies at the intersection of many fields of science of engineering, from psychology and economics to computer science and mathematics. Optimal control theory, game theory, the reward system of the brain and understanding animal behavior all address aspects of sequential learning, which concerns itself with the question: What is the optimal way to make decisions? I will discuss how this question led us to incorporate Bayesian Global Optimization (BGO) into materials design. Even though it has been applied to many materials systems, using manual and automated synthesis and characterization, there a number of problems with the method, which I will outline. I will then show how the sequential learning algorithm, Reinforcement Learning (RL), can overcome some of the issues, and will present how it can be applied to the compositional design of alloys.
Key words：Sequential design；Bayesian Global Optimization；Reinforcement learning；
Turab Lookman had worked in Los Almos National Laboratory for about 20 years. He has done intensive research in the areas of condensed matter physics and computational materials science. His authored and co-authored publications have been cited more than 14000 times with an h-index 59.