Seminar by Soojung Yang: Physics based learning for rare events
© 2024 EPFL
Join us for a seminar by Soojung Yang on Tuesday, August 13, at 2:00 PM in MXF 312. She is a PhD student focusing on conformational sampling of proteins via MD and ML in the group of Rafael Gomez-Bombarelli at MIT, and currently interning with Frank Noe at Microsoft Research Berlin.
Soojung will be talking about how Physics-Inspired Synthetic Data Augmentation Improves Collective Variable Learning for Rare Events.
In molecular dynamics simulations, rare events, such as protein folding, are typically studied using enhanced sampling techniques, most of which are based on the definition of a collective variable (CV) along which acceleration occurs. Obtaining an expressive CV is crucial, but often hindered by the lack of information about the particular event, e.g., the transition from unfolded to folded conformation. Soojung and coworkes propose a simulation-free data augmentation strategy using physics-inspired metrics to generate geodesic interpolations resembling protein folding transitions, thereby improving sampling efficiency without true transition state samples. This new data can be used to improve the accuracy of CV-based enhanced sampling.