(Non)-penalized multilevel methods for non-uniformly log-concave distributions

DOI10.1214/24-EJP1099MaRDI QIDQ6126986zbMATH OpenOpenAlexFDO

Authors Maxime Egéa

Publication date 10 April 2024

Published in Electronic Journal of Probability (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2301.09471

zbMATH Keywords

ergodic diffusionEuler schemeLangevin algorithmmultilevelweakly convex

Mathematics Subject Classification ID

Computational methods for ergodic theory (approximation of invariant measures, computation of Lyapunov exponents, entropy, etc.) (37M25) Probabilistic models, generic numerical methods in probability and statistics (65C20)

Abstract: We study and develop multilevel methods for the numerical approximation of a log-concave probability $pi$ on $mathbb{R}^d$, based on (over-damped) Langevin diffusion. In the continuity of cite{art:egeapanloup2021multilevel} concentrated on the uniformly log-concave setting, we here study the procedure in the absence of the uniformity assumption. More precisely, we first adapt an idea of cite{art:DalalyanRiouKaragulyan} by adding a penalization term to the potential to recover the uniformly convex setting. Such approach leads to an extit{$varepsilon$-complexity} of the order $varepsilo{n}^{-5}pi(|.{|}^2{)}^{3}d$ (up to logarithmic terms). Then, in the spirit of cite{art:gadat2020cost}, we propose to explore the robustness of the method in a weakly convex parametric setting where the lowest eigenvalue of the Hessian of the potential $U$ is controlled by the function $U(x{)}^{-r}$ for $rin(0,1)$. In this intermediary framework between the strongly convex setting ($r=0$) and the ``Laplace case ($r=1$), we show that with the help of the control of exponential moments of the Euler scheme, we can adapt some fundamental properties for the efficiency of the method. In the ``best setting where $U$ is ${mathcalC}^3$ and $U(x{)}^{-r}$ control the largest eigenvalue of the Hessian, we obtain an $varepsilon$-complexity of the order $c_{ho,delta}varepsilo{n}^{-2-ho}{d}^{1+fracho2+(4-ho+delta)r}$ for any $ho>0$ (but with a constant $c_{ho,delta}$ which increases when $ho$ and $delta$ go to $0$).

Recommendations

• Sampling from a log-concave distribution with projected Langevin Monte Carlo
• Kinetic Langevin MCMC sampling without gradient Lipschitz continuity -- the strongly convex case
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
• Analysis of Langevin Monte Carlo via convex optimization

Cites work

• A general method for debiasing a Monte Carlo estimator
• A simple proof of the Poincaré inequality for a large class of probability measures
• Approaching the Kannan-Lovász-Simonovits and variance conjectures
• Complexity analysis of accelerated MCMC methods for Bayesian inversion
• Exact estimation for Markov chain equilibrium expectations
• Exponential convergence of Langevin distributions and their discrete approximations
• General multilevel adaptations for stochastic approximation algorithms of Robbins-Monro and Polyak-Ruppert type
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• scientific article; zbMATH DE number 51724 (Why is no real title available?)
• scientific article; zbMATH DE number 2000348 (Why is no real title available?)
• Improved bounds for discretization of Langevin diffusions: near-optimal rates without convexity
• Markov chains and stochastic stability
• Multi-level Monte Carlo methods for the approximation of invariant measures of stochastic differential equations
• Multi-level stochastic approximation algorithms
• Multilevel Monte Carlo Path Simulation
• Recursive computation of the invariant distribution of a diffusion
• RECURSIVE COMPUTATION OF THE INVARIANT DISTRIBUTION OF A DIFFUSION: THE CASE OF A WEAKLY MEAN REVERTING DRIFT
• Second-order discretization schemes of stochastic differential systems for the computation of the invariant law
• Self-improvement of the Bakry-Emery criterion for Poincaré inequalities and Wasserstein contraction using variable curvature bounds
• Statistical Romberg extrapolation: a new variance reduction method and applications to option pricing
• Stochastic stability of differential equations. With contributions by G. N. Milstein and M. B. Nevelson
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
• Unadjusted Langevin algorithm with multiplicative noise: total variation and Wasserstein bounds
• Unbiased estimators and multilevel Monte Carlo
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
• Weighted Csiszár-Kullback-Pinsker inequalities and applications to transportation inequalities
• Weighted multilevel Langevin simulation of invariant measures

This page was built for publication: (Non)-penalized multilevel methods for non-uniformly log-concave distributions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6126986)