- This event has passed.
Machine Learning, Uncertainty Quantification, and Differentiable Optimization for Accelerating Calculations in Catalysis
July 28, 2020 from 10:00 am — 11:00 am
Zoom Webinar with Zachary Ulissi
Department of Chemical Engineering
Carnegie Mellon University, Pittsburgh, PA
I will discuss several recent efforts to make machine learning methods for molecular simulations more robust and applicable to day-to-day use in computational chemistry. First I will discuss recent improvements in making active learning methods robust for individual calculations (developed as the open source AMP package). I will then discuss methods for quantifying uncertainty in property predictions using graph convolution models to learn features as inputs to Gaussian Process models. Finally I will discuss recent work to directly predict relaxed state structures for adsorbates by directly embedding the relaxations into the machine learning model during the fitting process. All of these methods are applicable to other atomistic simulations.
Zack Ulissi joined Carnegie Mellon University in 2017, after doing his BS/BE at the University of Delaware, MASt at Cambridge, PhD at MIT and post-doc at Stanford. His research at MIT focused on the applications of systems engineering methods to understanding selective nanoscale carbon nanotube devices and sensors under the supervision of Michael Strano and Richard Braatz. Prof. Ulissi did his postdoctoral work at Stanford with Jens Nørskov where he worked on machine learning techniques to simplify complex catalyst reaction networks, applied to the electrochemical reduction of N2 and CO2 to fuels. The Ulissi group builds on this foundation to model, understand, and design nanoscale interfaces using machine learning and predictive methods to guide detailed molecular simulations. Recent awards include the ACS Petroleum Research Fund Doctoral New Investigator Award and the 3M Non-Tenured Faculty Award.
For Zoom information please contact email@example.com