English

Rapid Convergence of the Unadjusted Langevin Algorithm: Isoperimetry Suffices

Data Structures and Algorithms 2022-03-04 v4 Machine Learning Probability Machine Learning

Abstract

We study the Unadjusted Langevin Algorithm (ULA) for sampling from a probability distribution ν=ef\nu = e^{-f} on Rn\mathbb{R}^n. We prove a convergence guarantee in Kullback-Leibler (KL) divergence assuming ν\nu satisfies a log-Sobolev inequality and the Hessian of ff is bounded. Notably, we do not assume convexity or bounds on higher derivatives. We also prove convergence guarantees in R\'enyi divergence of order q>1q > 1 assuming the limit of ULA satisfies either the log-Sobolev or Poincar\'e inequality. We also prove a bound on the bias of the limiting distribution of ULA assuming third-order smoothness of ff, without requiring isoperimetry.

Keywords

Cite

@article{arxiv.1903.08568,
  title  = {Rapid Convergence of the Unadjusted Langevin Algorithm: Isoperimetry Suffices},
  author = {Santosh S. Vempala and Andre Wibisono},
  journal= {arXiv preprint arXiv:1903.08568},
  year   = {2022}
}

Comments

v4: Updated discussion and added properties of biased limit v3: Simplified analysis of R\'enyi divergence, improved exposition, and added figures v2: Added analysis of R\'enyi divergence and Poincar\'e assumption