The identification of meaningful collective coordinates plays a key role in the study of complex molecular systems whose essential dynamics are characterized by rare or slow transition events. In a recent years, precise defining characteristics of such coordinates were identified and linked to the existence of a so-called transition manifold. This theory gives rise to a novel learning methods for the identification of transition manifolds that relies on short parallel MD simulations only and yields approximation of the long time behavior of the system including accuracy guarantees.
After presenting these results, the problem of learning the effective dynamics on the transition manifold will be discussed and first theoretical results as well as algorithmic approaches will be outlined.
Back to Workshop III: Validation and Guarantees in Learning Physical Models: from Patterns to Governing Equations to Laws of Nature