Scientific Uses of Automatic Differentiation - DDPS

Explore the scientific applications of automatic differentiation in this 1-hour 8-minute lecture by Michael Brenner, part of the Data-Driven Physical Simulations series. Discover how tools underlying the machine learning revolution, particularly automatic differentiation, offer significant opportunities for scientific discovery. Learn about various research directions utilizing automatic differentiation and large-scale optimization to solve scientific problems, including developing new algorithms for partial differential equations, designing energy landscapes for self-assembly, uncovering unstable solutions in fluid dynamics, modeling organismal development, implementing nonequilibrium statistical mechanics protocols, designing fluid rheology, and applying statistical mechanics algorithms to protein self-assembly. Gain insights into innovative approaches and thought processes for leveraging these tools in scientific research.