Speaker: Daniel McBride, UTK

Title: Structure-preserving numerical integration for Hamiltonian systems.

Abstract: Hamiltonian systems are a classical object of study in physics and have become, in recent decades, important in certain high-dimensional statistical inference methods. Exact flows for solutions of Hamilton's equations are symplectic and time-reversible, so it is important in many applications to use numerical integration schemes that preserve this structure. In this talk, we will see some first examples, and discuss under what conditions a scheme is structure-preserving in this sense.