ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1805.01648
13
166

Sharp convergence rates for Langevin dynamics in the nonconvex setting

4 May 2018
Xiang Cheng
Niladri S. Chatterji
Yasin Abbasi-Yadkori
Peter L. Bartlett
Michael I. Jordan
ArXivPDFHTML
Abstract

We study the problem of sampling from a distribution p∗(x)∝exp⁡(−U(x))p^*(x) \propto \exp\left(-U(x)\right)p∗(x)∝exp(−U(x)), where the function UUU is LLL-smooth everywhere and mmm-strongly convex outside a ball of radius RRR, but potentially nonconvex inside this ball. We study both overdamped and underdamped Langevin MCMC and establish upper bounds on the number of steps required to obtain a sample from a distribution that is within ϵ\epsilonϵ of p∗p^*p∗ in 111-Wasserstein distance. For the first-order method (overdamped Langevin MCMC), the iteration complexity is O~(ecLR2d/ϵ2)\tilde{\mathcal{O}}\left(e^{cLR^2}d/\epsilon^2\right)O~(ecLR2d/ϵ2), where ddd is the dimension of the underlying space. For the second-order method (underdamped Langevin MCMC), the iteration complexity is O~(ecLR2d/ϵ)\tilde{\mathcal{O}}\left(e^{cLR^2}\sqrt{d}/\epsilon\right)O~(ecLR2d​/ϵ) for an explicit positive constant ccc. Surprisingly, the iteration complexity for both these algorithms is only polynomial in the dimension ddd and the target accuracy ϵ\epsilonϵ. It is exponential, however, in the problem parameter LR2LR^2LR2, which is a measure of non-log-concavity of the target distribution.

View on arXiv
Comments on this paper