A team led by researchers at the University of California, San Diego (UC San Diego) and the San Diego Supercomputer Center (SDSC) used machine learning techniques to develop the most accurate models to date for simulations of water.
The team also included researchers at the École Polytechnique Fédérale de Lausanne in Switzerland, Cambridge University in the U.K., and the University of Göttingen in Germany.
The work highlights how popular machine learning techniques can be used to construct predictive molecular models, including other "generic" molecules based on quantum mechanical reference data.
To represent many-body interactions in water, the researchers investigated the performance of three machine learning techniques: permutationally invariant polynomials, neural networks, and Gaussian approximation potentials.
"We have demonstrated that these different machine learning techniques can effectively be employed to encode the highly complex quantum mechanical many-body interactions that arise when molecules interact," says UC San Diego's Thuong Nguyen.
The complex neural networks with associated optimization processes were developed on SDSC's Comet supercomputer GPU resources and Maverick, based at the Texas Advanced Computing Center (TACC), with allocations provided by the eXtreme Science and Engineering Discovery Environment (XSEDE).
From UC San Diego News Center
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found