Machine Learning Techniques for Photocatalysis and Materials Discovery

Material science and engineering research has long utilized the trial-and-error experimentation. However, the approach is costly and time-consuming. This thesis intends to introduce machine learning as an efficient tool in alleviating the burden associated with the conventional approaches in material science research. We first discuss how discriminative models can be incorporated to facilitate the photocatalysis process and reduce the experimental burden. We then present a generative model framework for inorganic material discovery with desired properties. Finally, we extend and enhance the generative model with improved loss function to enable more effective material generation.