Non-reversible Markov chain Monte Carlo and hypocoercivityCRC 1720: Early Career Seminar

by Francis Lörler

Tuesday Jan 27, 2026, 2:15 PM → 3:15 PM Europe/Berlin

Endenicher Allee 60/1-016 - Lipschitzsaal (Mathezentrum)

How can one construct a Markov process that converges to a prescribed invariant measure as fast as possible? This question lies at the heart of Markov chain Monte Carlo, a cornerstone of modern computational mathematics. We will explore how non-reversibility and degenerate noise can overcome the slow converge of traditional reversible algorithms. We then show how these ideas are connected with the theory of hypocoercivity, which explains rapid convergence in the absence of direct dissipation, through the concept of second-order lifts.