Machine learning of molecular potential energy surfaces (PESs) falls within three broad categories: Permutationally Invariant Polynomials (PIPs), Gaussian Process Regression and Neural Networks. We introduced PIPs almost 15 years ago, using to determine a global PES for the highly fluxional cation CH5+. More recently, that approach has been used by us and others to develop many-body PESs for water and protonated water clusters, and several hydrate clathrates. I will describe the PIP methodology and illustrate it for several many-body PESs.

Ab initio many-body potentials from “machine learning” before we knew that terminology

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.