Higher order Langevin Monte Carlo algorithm

DOI10.1214/19-EJS1615MaRDI QIDQ2326072zbMATH OpenFDO

Authors Sotirios Sabanis, Ying Zhang

Publication date 4 October 2019

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

Full work available at URL https://arxiv.org/abs/1808.00728, https://projecteuclid.org/euclid.ejs/1570068044

zbMATH Keywords

higher order algorithmmachine learningMarkov chain Monte Carlorate of convergencesampling problemsuper-linear coefficients

Mathematics Subject Classification ID

Computational methods for problems pertaining to probability theory (60-08) Stochastic ordinary differential equations (aspects of stochastic analysis) (60H10) Monte Carlo methods (65C05)

Abstract: A new (unadjusted) Langevin Monte Carlo (LMC) algorithm with improved rates in total variation and in Wasserstein distance is presented. All these are obtained in the context of sampling from a target distribution $pi$ that has a density $hatpi$ on $mathbb{R}^d$ known up to a normalizing constant. Moreover, $-loghatpi$ is assumed to have a locally Lipschitz gradient and its third derivative is locally H"{o}lder continuous with exponent $$. Non-asymptotic bounds are obtained for the convergence to stationarity of the new sampling method with convergence rate $$ in Wasserstein distance, while it is shown that the rate is 1 in total variation even in the absence of convexity. Finally, in the case where $-loghatpi$ is strongly convex and its gradient is Lipschitz continuous, explicit constants are provided.

Recommendations

• Analysis of Langevin Monte Carlo via convex optimization
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• Sampling from a log-concave distribution with projected Langevin Monte Carlo
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient

Cites work

• A note on tamed Euler approximations
• Analysis and geometry of Markov diffusion operators
• Ergodicity for SDEs and approximations: locally Lipschitz vector fields and degenerate noise.
• Euler approximations with varying coefficients: the case of superlinearly growing diffusion coefficients
• Evaluating Derivatives
• Evaluating higher derivative tensors by forward propagation of univariate Taylor series
• Exponential convergence of Langevin distributions and their discrete approximations
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• scientific article; zbMATH DE number 3839062 (Why is no real title available?)
• scientific article; zbMATH DE number 54145 (Why is no real title available?)
• scientific article; zbMATH DE number 1149166 (Why is no real title available?)
• scientific article; zbMATH DE number 1546853 (Why is no real title available?)
• scientific article; zbMATH DE number 1405932 (Why is no real title available?)
• Introductory lectures on convex optimization. A basic course.
• Markov Chains and Stochastic Stability
• Non-negative matrix factorization with sparseness constraints
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• On explicit order 1.5 approximations with varying coefficients: the case of super-linear diffusion coefficients
• On Milstein approximations with varying coefficients: the case of super-linear diffusion coefficients
• Online learning for matrix factorization and sparse coding
• Optimal Transport
• Some bounds on the complexity of gradients, Jacobians, and Hessians
• Stability of Markovian processes III: Foster–Lyapunov criteria for continuous-time processes
• Stochastic C-stability and B-consistency of explicit and implicit Milstein-type schemes
• Strong and weak divergence in finite time of Euler's method for stochastic differential equations with non-globally Lipschitz continuous coefficients
• Strong convergence of an explicit numerical method for SDEs with nonglobally Lipschitz continuous coefficients
• The tamed Milstein method for commutative stochastic differential equations with non-globally Lipschitz continuous coefficients
• The tamed unadjusted Langevin algorithm
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient

Cited in

(27)

• Normalizing constants of log-concave densities
• Is there an analog of Nesterov acceleration for gradient-based MCMC?
• Sampling from non-smooth distributions through Langevin diffusion
• Unadjusted Langevin algorithm for sampling a mixture of weakly smooth potentials
• Constrained ensemble Langevin Monte Carlo
• Error estimates of the backward Euler-Maruyama method for multi-valued stochastic differential equations
• On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case
• The tamed unadjusted Langevin algorithm
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• Langevin diffusions and the Metropolis-adjusted Langevin algorithm
• Nonasymptotic estimates for stochastic gradient Langevin dynamics under local conditions in nonconvex optimization
• Convergence of Langevin MCMC in KL-divergence
• High-order Langevin diffusion yields an accelerated MCMC algorithm
• Sublinear convergence of a tamed stochastic gradient descent method in Hilbert space
• On Stochastic Gradient Langevin Dynamics with Dependent Data Streams: The Fully Nonconvex Case
• Analysis of Langevin Monte Carlo via convex optimization
• Gradient-based adaptive importance samplers
• Unadjusted Langevin algorithm with multiplicative noise: total variation and Wasserstein bounds
• Taming Neural Networks with TUSLA: Nonconvex Learning via Adaptive Stochastic Gradient Langevin Algorithms
• Lévy Langevin Monte Carlo
• Kinetic Langevin MCMC sampling without gradient Lipschitz continuity -- the strongly convex case
• Non-asymptotic convergence bounds for modified tamed unadjusted Langevin algorithm in non-convex setting
• Noise-free sampling algorithms via regularized Wasserstein proximals
• Taming the interacting particle Langevin algorithm: the superlinear case
• Taming under isoperimetry

This page was built for publication: Higher order Langevin Monte Carlo algorithm

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2326072)