“Low-dimensional representations for learning fast molecular dynamics simulations and new visualizations”

Protein folding and other molecular dynamics simulations are a crucial tool for understanding the biochemical underpinnings of life, but even with powerful computers researchers are unable to simulate the important state transitions at physiological temperatures. I will present some new work which uses local dimensionality reduction to learn approximate simulations which run hundreds of times faster than the originals, along with some web-based visualizations we’re developing to explore the results.
Friday, February 7
Noon – 1pm
D106 LSRC (Campus map)