User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
From MaRDI portal
Publication:2280028
Abstract: In this paper, we study the problem of sampling from a given probability density function that is known to be smooth and strongly log-concave. We analyze several methods of approximate sampling based on discretizations of the (highly overdamped) Langevin diffusion and establish guarantees on its error measured in the Wasserstein-2 distance. Our guarantees improve or extend the state-of-the-art results in three directions. First, we provide an upper bound on the error of the first-order Langevin Monte Carlo (LMC) algorithm with optimized varying step-size. This result has the advantage of being horizon free (we do not need to know in advance the target precision) and to improve by a logarithmic factor the corresponding result for the constant step-size. Second, we study the case where accurate evaluations of the gradient of the log-density are unavailable, but one can have access to approximations of the aforementioned gradient. In such a situation, we consider both deterministic and stochastic approximations of the gradient and provide an upper bound on the sampling error of the first-order LMC that quantifies the impact of the gradient evaluation inaccuracies. Third, we establish upper bounds for two versions of the second-order LMC, which leverage the Hessian of the log-density. We nonasymptotic guarantees on the sampling error of these second-order LMCs. These guarantees reveal that the second-order LMC algorithms improve on the first-order LMC in ill-conditioned settings.
Recommendations
- Higher order Langevin Monte Carlo algorithm
- Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
- On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case
- Analysis of Langevin Monte Carlo via convex optimization
- Sampling from a log-concave distribution with projected Langevin Monte Carlo
Cites work
- A variational perspective on accelerated methods in optimization
- An introduction to optimization
- Consistency and fluctuations for stochastic gradient Langevin dynamics
- Convergence of Langevin MCMC in KL-divergence
- Criteria for recurrence and existence of invariant measures for multidimensional diffusions
- Efficient Bayesian computation by proximal Markov chain Monte Carlo: when Langevin meets Moreau
- Exponential convergence of Langevin distributions and their discrete approximations
- Geometric ergodicity of Metropolis algorithms
- High-dimensional Bayesian inference via the unadjusted Langevin algorithm
- scientific article; zbMATH DE number 1022658 (Why is no real title available?)
- Introductory lectures on convex optimization. A basic course.
- Langevin diffusions and Metropolis-Hastings algorithms
- Langevin-type models II: Self-targeting candidates for MCMC algorithms
- Langevin-type models. I: Diffusions with given stationary distributions and their discretizations
- Noisy Monte Carlo: convergence of Markov chains with approximate transition kernels
- Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
- Normalizing constants of log-concave densities
- Optimal Scaling of Discrete Approximations to Langevin Diffusions
- Optimal transport bounds between the time-marginals of a multidimensional diffusion and its Euler scheme
- Pathwise optimal transport bounds between a one-dimensional diffusion and its Euler scheme
- Sampling can be faster than optimization
- Sampling from a log-concave distribution with projected Langevin Monte Carlo
- Sampling from non-smooth distributions through Langevin diffusion
- Some bounds on the complexity of gradients, Jacobians, and Hessians
- Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
Cited in
(83)- Is there an analog of Nesterov acceleration for gradient-based MCMC?
- Unadjusted Langevin algorithm for sampling a mixture of weakly smooth potentials
- Oracle lower bounds for stochastic gradient sampling algorithms
- Improved bounds for discretization of Langevin diffusions: near-optimal rates without convexity
- Stochastic zeroth-order discretizations of Langevin diffusions for Bayesian inference
- Stochastic gradient Hamiltonian Monte Carlo for non-convex learning
- Constrained ensemble Langevin Monte Carlo
- On sampling from a log-concave density using kinetic Langevin diffusions
- Optimal scaling of random-walk Metropolis algorithms on general target distributions
- On stochastic gradient Langevin dynamics with dependent data streams in the logconcave case
- Higher order Langevin Monte Carlo algorithm
- Control variates for stochastic gradient MCMC
- Quantitative contraction rates for Markov chains on general state spaces
- Nonasymptotic bounds for sampling algorithms without log-concavity
- Nonasymptotic estimates for stochastic gradient Langevin dynamics under local conditions in nonconvex optimization
- Stochastic gradient Langevin dynamics with adaptive drifts
- Variance reduction for dependent sequences with applications to stochastic gradient MCMC
- High-order Langevin diffusion yields an accelerated MCMC algorithm
- Stochastic gradient MCMC for state space models
- scientific article; zbMATH DE number 7626720 (Why is no real title available?)
- scientific article; zbMATH DE number 7626754 (Why is no real title available?)
- scientific article; zbMATH DE number 7626757 (Why is no real title available?)
- Global convergence of stochastic gradient Hamiltonian Monte Carlo for nonconvex stochastic optimization: nonasymptotic performance bounds and momentum-based acceleration
- Maximum likelihood estimation of regularization parameters in high-dimensional inverse problems: an empirical Bayesian approach. II: Theoretical analysis
- Asymptotic analysis via stochastic differential equations of gradient descent algorithms in statistical and computational paradigms
- scientific article; zbMATH DE number 7415079 (Why is no real title available?)
- Replica exchange for non-convex optimization
- On Stochastic Gradient Langevin Dynamics with Dependent Data Streams: The Fully Nonconvex Case
- Analysis of Langevin Monte Carlo via convex optimization
- Laplacian smoothing stochastic gradient Markov chain Monte Carlo
- Stochastic gradient Markov chain Monte Carlo
- Ensemble Kalman sampler: mean-field limit and convergence analysis
- Optimal convergence rate of Hamiltonian Monte Carlo for strongly logconcave distributions
- Statistical Finite Elements via Langevin Dynamics
- Global Optimization via Schrödinger–Föllmer Diffusion
- Smoothing unadjusted Langevin algorithms for nonsmooth composite potential functions
- Ergodicity of supercritical SDEs driven by \(\alpha \)-stable processes and heavy-tailed sampling
- (Non)-penalized multilevel methods for non-uniformly log-concave distributions
- A fresh Take on ‘Barker Dynamics’ for MCMC
- Improving sampling accuracy of stochastic gradient MCMC methods via non-uniform subsampling of gradients
- Taming Neural Networks with TUSLA: Nonconvex Learning via Adaptive Stochastic Gradient Langevin Algorithms
- Multi-index antithetic stochastic gradient algorithm
- The forward-backward envelope for sampling with the overdamped Langevin algorithm
- NF-ULA: normalizing flow-based unadjusted Langevin algorithm for imaging inverse problems
- Accelerated Bayesian imaging by relaxed proximal-point Langevin sampling
- Augmented projection Wasserstein distances: multi-dimensional projection with neural surface
- Kinetic Langevin MCMC sampling without gradient Lipschitz continuity -- the strongly convex case
- Proximal Langevin sampling with inexact proximal mapping
- Riemannian Langevin algorithm for solving semidefinite programs
- Non-asymptotic convergence bounds for modified tamed unadjusted Langevin algorithm in non-convex setting
- Contraction rate estimates of stochastic gradient kinetic Langevin integrators
- Conformalized-DeepONet: a distribution-free framework for uncertainty quantification in deep operator networks
- On high-dimensional classification by sparse generalized Bayesian logistic regression
- A tamed Euler scheme for SDEs with non-locally integrable drift coefficient
- Gradient-adjusted underdamped Langevin dynamics for sampling
- Parallelized midpoint randomization for Langevin Monte Carlo
- Statistical finite elements via interacting particle Langevin dynamics
- A splitting Hamiltonian Monte Carlo method for efficient sampling
- Unadjusted Langevin algorithm for non-convex weakly smooth potentials
- Resolving the mixing time of the Langevin algorithm to its stationary distribution for log-concave sampling
- Batch normalization preconditioning for stochastic gradient Langevin dynamics
- Interacting particle Langevin algorithm for maximum marginal likelihood estimation
- Bayesian Federated Learning with Hamiltonian Monte Carlo: Algorithm and Theory
- Wasserstein convergence guarantees for a general class of score-based generative models
- Concentration properties of fractional posterior in 1-bit matrix completion
- HMC and underdamped Langevin united in the unadjusted convex smooth case
- Query lower bounds for log-concave sampling
- Bayesian analysis of exponential random graph models using stochastic gradient Markov chain Monte Carlo
- Stochastic gradient MCMC for nonlinear state space models
- Stochastic gradient Langevin dynamics for (weakly) log-concave posterior distributions
- Discrete sticky couplings of functional autoregressive processes
- Wasserstein mixing time of the unadjusted Langevin algorithm
- Score-based deterministic density sampling
- Estimates of the numerical density for stochastic differential equations with multiplicative noise
- (De)-regularized maximum mean discrepancy gradient flow
- Targeted stochastic gradient MCMC for HMMs with rare latent states
- Subspace Langevin Monte Carlo
- Multimodal sampling via Schrödinger-Föllmer samplers with temperatures
- On the computational cost of Stochastic Gradient Langevin Dynamics
- Theoretical guarantees for stochastic gradient Langevin dynamics
- Fast replica exchange stochastic gradient Langevin dynamics
- Kinetic interacting particle Langevin Monte Carlo
- Piecewise deterministic sampling for constrained distributions
This page was built for publication: User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2280028)
