Deep Learning for Multi-Scale Molecular Modelling

Deep learning has emerged as a promising tool for a variety of applications
in scientific modeling.
However, while deep learning is a powerful tool for fitting data,
constructing reliable and practical deep learning-based physical models
is still a very non-trivial task.
In this talk, we discuss the important issues and illustrate the relevant
ideas in the context of constructing
potential energy and free energy surfaces for molecular modeling.
In particular, we will discuss the Deep Potential and Deep Potential Generator
scheme and the
Reinforced Dynamics for enhanced sampling and free energy calculation.

~