Explore the challenges and advancements in developing accurate global machine learning force fields (MLFFs) for large molecular systems in this 45-minute conference talk. Delve into the limitations of current MLFFs, which struggle to scale beyond a few dozen atoms due to increasing model complexity. Examine the drawbacks of introducing locality assumptions for larger molecules, including poor representation of non-local interactions and potential inaccuracies in molecular dynamics simulations. Discover research directions aimed at reconstructing precise global MLFFs for systems with up to several hundred atoms, without relying on localization of atomic interactions or other potentially uncontrolled approximations. Gain insights into the latest approaches for capturing collective many-atom interactions in complex molecular structures.
Accurate Global Machine Learning Force Fields for Molecules with Hundreds of Atoms
Institute for Pure & Applied Mathematics (IPAM) via YouTube
Overview
Syllabus
Stefan Chmiela - Accurate global machine learning force fields for molecules with hundreds of atoms
Taught by
Institute for Pure & Applied Mathematics (IPAM)
Related Courses
-
Non-Locality in Machine Learning Force Fields - IPAM at UCLA
-
Benchmark and Critical Evaluation for ML Force Fields with Molecular Simulations
-
Uncertainty-Aware Machine Learning Models of Many-Body Atomic Interactions
-
Advancing Molecular Simulation with Deep Learning - IPAM at UCLA
-
Extending the Reach of Quantum Monte Carlo Methods via Machine Learning
-
Acceleration and Analysis of Molecular Dynamics Simulation Using Machine Learning
