The journal of physical chemistry. A
Investigation of Data Set Portability on Various Machine-Learned Interaction Potentials for Pyrophyllite Clay.
Chloe Sanz, Colin Bousige, Pierre Mignon
Published: 202610.1021/acs.jpca.5c07319
Abstract
The influence of the architecture and descriptors on the accuracy of a neural network potential is investigated. Particularly, we focused on high-dimension machine-learned interaction potentials (MLIPs) using descriptors based on atom-centered symmet…
Preview only. Read the full abstract at the source