Explore the current state of neural network interatomic potentials (NNIPs) in a 41-minute conference talk presented by Justin Smith of NVIDIA at IPAM's workshop for Co-design for the Exascale and IPAM Hackathon. Delve into the world of atomistic simulation and its significance in biochemistry, material science, and reactive chemistry research. Compare traditional approaches of classical potentials and ab initio quantum chemistry with the emerging field of NNIPs. Examine the progress made by NNIPs in bridging the gap between speed and accuracy, while also addressing the challenges hindering their widespread adoption. Gain insights into ideal training data set generation, model validation, bespoke vs. general models, computational speed comparisons, and the need for accessible workflows for non-experts. Discover ongoing research and software engineering efforts aimed at overcoming these obstacles and advancing the field of NNIPs.
The State of Neural Network Interatomic Potentials - IPAM at UCLA
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Syllabus
Justin Smith - The state of neural network interatomic potentials - IPAM at UCLA
Taught by
Institute for Pure & Applied Mathematics (IPAM)