Hadrien Vroylandt (Institut des Sciences du Calcul et des Données, Sorbonne Université)
Learning the dynamics of systems with memory : Generalized Langevin equations
Generalized Langevin equations with non-linear forces and memory kernels are commonly used to describe the effective dynamics of coarse-grained variables in molecular dynamics. Such reduced dynamics play an essential role in the study of a broad class of processes, ranging from chemical reactions in solution to conformational changes in biomolecules or phase transitions in condensed matter systems. I will first discuss the derivation of the generalized Langevin equations, emphasizing the need for memory in the effective dynamics due to the lack of a proper separation of time scales. Then, I will turn on the inference of such generalized Langevin equations from observed trajectories, using a maximum likelihood approach. This data-driven approach provides a reduced dynamical model for collective variables, enabling the accurate sampling of their long-time dynamical properties at a computational cost drastically reduced with respect to all-atom numerical simulations. I will illustrate the potential of this method on several model systems, both in and out of equilibrium.
En visioconférence par zoom
Lien : https://us06web.zoom.us/j/81533150574?pwd=NlM1ZlJvUTQvZUxpVmV6QWdMbmlLQT09
Meeting ID: 815 3315 0574