English

High-Order Langevin Monte Carlo Algorithms

Machine Learning 2025-08-26 v1 Machine Learning Probability

Abstract

Langevin algorithms are popular Markov chain Monte Carlo (MCMC) methods for large-scale sampling problems that often arise in data science. We propose Monte Carlo algorithms based on the discretizations of PP-th order Langevin dynamics for any P3P\geq 3. Our design of PP-th order Langevin Monte Carlo (LMC) algorithms is by combining splitting and accurate integration methods. We obtain Wasserstein convergence guarantees for sampling from distributions with log-concave and smooth densities. Specifically, the mixing time of the PP-th order LMC algorithm scales as O(d1R/ϵ12R)O\left(d^{\frac{1}{R}}/\epsilon^{\frac{1}{2R}}\right) for R=41{P=3}+(2P1)1{P4}R=4\cdot 1_{\{ P=3\}}+ (2P-1)\cdot 1_{\{ P\geq 4\}}, which has a better dependence on the dimension dd and the accuracy level ϵ\epsilon as PP grows. Numerical experiments illustrate the efficiency of our proposed algorithms.

Keywords

Cite

@article{arxiv.2508.17545,
  title  = {High-Order Langevin Monte Carlo Algorithms},
  author = {Thanh Dang and Mert Gurbuzbalaban and Mohammad Rafiqul Islam and Nian Yao and Lingjiong Zhu},
  journal= {arXiv preprint arXiv:2508.17545},
  year   = {2025}
}

Comments

73 pages, 3 figures, 1 table

R2 v1 2026-07-01T05:03:46.730Z