University of California, Los Angeles,
Keywords: glass, machine learning, modeling
Summary:Unlike crystalline materials, glasses can virtually feature any composition and stoichiometry, which creates limitless opportunities to develop new glass formulations with unusual properties and functionalities. However, this large compositional space renders traditional Edisonian trial-and-error discovery approaches poorly efficient (“curse of dimensionality”). In this presentation, I will present some of our recent effort in combining machine learning and simulations to inform, optimize, and accelerate glass manufacturing.