University of Waterloo

Published purpose

This Project will develop new machine learning algorithms for automating aspects of physical chemistry by discovering new pathways for creating materials with desired properties. The goal is to use the framework of reinforcement learning as a means of discovering, controlling, and optimizing the production of new materials. This Project will explore a novel new direction to perform decision making associated with established but highly complex processes used to transform one chemical compound into another. From an algorithmic point of view, this research will tackle the challenge of working with highly nonlinear relationships and generating simulation data that is of high quality.