Skip to main content

Seminars - Probability and statistics

KTH's seminar series on probability, statistics, data science and related topics.

The seminar takes place on Mondays 15:15–16:16 in room 3721 (7th floor).

For those that prefer to participate remotely, you can follow the seminar via Zoom, meeting ID 62144698204

Next Talk

Monday, December 6, 15:15
Nawaf Bou-Rabee (Rutgers)
Upper Bounds for Mixing Time of Unadjusted Hamiltonian Monte Carlo by Couplings

Hamiltonian Monte Carlo (HMC) is a Markov Chain Monte Carlo method that is frequently used in statistical physics, statistics, and machine learning. The transition step of HMC uses a combination of Hamiltonian dynamics and velocity randomizations. The Hamiltonian dynamics is typically discretized using a reversible/volume-preserving integrator, and the discretization bias can either be borne (unadjusted HMC) or eliminated by Metropolis-Hastings (adjusted HMC). Despite its empirical success, until a few years ago there were few mixing time guarantees for HMC. Now, approaches to quantify convergence to equilibrium based on coupling, conductance and hypocoercivity have been developed. I will talk about the coupling approach, and show how it can be used to obtain mixing time guarantees for unadjusted HMC in some high-dimensional and non-convex models. This talk is based on joint work with Andreas Eberle (Bonn) and Katharina Schuh (Bonn).