Homepage
Nawaf Bou-Rabee
Professor of Mathematics at Rutgers University-Camden, Visiting Scholar at the Flatiron Institute, and Adjunct Professor of Statistics and Data Science at the Wharton School.
I work on the convergence of Markov chain Monte Carlo, mostly through couplings: upper bounds on mixing times for Hamiltonian Monte Carlo, kinetic Langevin samplers and the No-U-Turn Sampler (NUTS), and lower bounds showing where those rates are sharp. Informed by these theoretical insights, I also design locally adaptive samplers, most notably WALNUTS, a generalization of NUTS that adapts the step size within each orbit. The framework behind it is described in our survey on Gibbs self-tuning (GIST).
From 2019 to 2021, I was a Visiting Professor in Probability Theory and Stochastic Analysis at Bonn University. I earned my PhD in Applied and Computational Mathematics from Caltech before holding an NSF Mathematical Sciences Postdoctoral Research Fellowship and a Courant Instructorship at NYU. My work has been supported by several NSF grants, and I’ve been fortunate to receive the Rutgers-Camden Chancellor’s Award for Outstanding Research and a Humboldt Research Fellowship for Experienced Researchers.
Curriculum Vitae · Publications · Google Scholar · arXiv · ORCID · GitHub · MathOverflow · LinkedIn
Recent
Preprint: Persistence Pays: Square-root condition-number dependence for Metropolis-adjusted kinetic Langevin Monte Carlo with a non-reversibly updated accept/reject uniform. arXiv · Lean formalization · Note · Andrew Gelman’s blog
Seminar: End-to-End Total Variation Guarantees for Sequence-Parallel Markov Chain Monte Carlo. Wharton Statistics and Data Science Seminar, October 2026. Abstract
Plenary lecture, NUTS for NUTS: Advances in No-U-Turn Sampling and the Future of Markov Chain Monte Carlo. SIAM Conference on Uncertainty Quantification, March 2026. Slides
WALNUTS, the No-U-Turn Sampler with adaptive step size. Journal of Machine Learning Research 27, 2026. Paper · Code · Interactive demo
Preprint: Provable Non-Acceleration of Standard Strang Splittings of Kinetic Langevin Dynamics. arXiv · Note
Preprint: On Couplings for Kinetic Langevin Diffusions, with Sonja Cox and Roy Schieven. arXiv · Note · Slides
Co-organizer, Stochastic Computation, FoCM 2026, University of Vienna, July 2026, with Mireille Bossy and David Cohen. Workshop
Talk: the reversibility and mixing time of the No-U-Turn Sampler, October 2024. Video