What type of study is this?

This is a Experimental Study study.

October 13, 2025Open Access

Contractive kinetic Langevin samplers beyond global Lipschitz continuity

ILIosif LytrasNational Technical University of Athens PMPanagiotis Mertikopoulos

Key Points

The proposed discretizations of the kinetic Langevin SDE are contractive and adhere to a log-Sobolev inequality, enhancing sampling methods.
Non-asymptotic bounds were established in 2-Wasserstein distance, ensuring better convergence to the target measure for both algorithms.
The taming scheme facilitates the effective handling of superlinear gradient growth in log-concave distributions, improving sampling efficiency.
Overall, the methods may provide robust solutions for sampling complex distributions in various applied settings.

Abstract

In this paper, we examine the problem of sampling from log-concave distributions with (possibly) superlinear gradient growth under kinetic (underdamped) Langevin algorithms. Using a carefully tailored taming scheme, we propose two novel discretizations of the kinetic Langevin SDE, and we show that they are both contractive and satisfy a log-Sobolev inequality. Building on this, we establish a series of non-asymptotic bounds in 2-Wasserstein distance between the law reached by each algorithm and the underlying target measure.

Perguntar à IA

Bookmark

View Full Paper