Atomic Cluster Expansion: A framework for fast and accurate ML force fields

Lennard-Jones Centre discussion group seminar by Dávid P. Kovács from the University of Cambridge.

This talk describes the Atomic Cluster Expansion (ACE) which provides a systematic framework to derive a formally complete set of symmetric polynomial basis functions. The talk shows that using the ACE features and linear regression it is possible to create highly accurate and fast force fields. The talk also shows some recent results of using equivariant ACE for the fitting of vectorial properties, such as dipole moments. Finally, the talk briefly presents multi-ACE, a framework unifying most existing machine learning force fields, including local models like SOAP -GAP, and the message passing neural networks.

The seminar was held on 25th April 2022.

🖥️ Check out our websites: https://linktr.ee/cumaterials