Analysis of Primal-Dual Langevin Algorithms

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Burger, Martin, Ehrhardt, Matthias J., Kuger, Lorenz, Weigand, Lukas
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910684415000576
author Burger, Martin
Ehrhardt, Matthias J.
Kuger, Lorenz
Weigand, Lukas
author_facet Burger, Martin
Ehrhardt, Matthias J.
Kuger, Lorenz
Weigand, Lukas
contents We analyze a recently proposed class of algorithms for the problem of sampling from probability distributions $μ^\ast$ in $\mathbb{R}^d$ with a Lebesgue density of the form $μ^\ast(x) \propto \exp(-f(Kx)-g(x))$, where $K$ is a linear operator and $f,g$ convex and non-smooth. The method is a generalization of the primal-dual hybrid gradient optimization algorithm to a sampling scheme. We give the iteration's continuous time limit, a stochastic differential equation in the joint primal-dual variable, and its mean field limit Fokker-Planck equation. Under mild conditions, the scheme converges to a unique stationary state in continuous and discrete time. Contrary to purely primal overdamped Langevin diffusion, the stationary state in continuous time does not have $μ^\ast$ as its primal marginal. Thus, further analysis is carried out to bound the bias induced by the partial dualization, and potentially correct for it in the diffusion. Time discretizations of the diffusion lead to implementable algorithms, but, as is typical in Langevin Monte Carlo methods, introduce further bias. We prove bounds for these discretization errors, which allow to give convergence results relating the produced samples to the target. We demonstrate our findings numerically first on small-scale examples in which we can exactly verify the theoretical results, and subsequently on typical examples of larger scale from Bayesian imaging inverse problems.
format Preprint
id arxiv_https___arxiv_org_abs_2405_18098
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Analysis of Primal-Dual Langevin Algorithms
Burger, Martin
Ehrhardt, Matthias J.
Kuger, Lorenz
Weigand, Lukas
Optimization and Control
35Q84, 47A52, 49N45, 60H10, 62F15, 65J22, 68U10
We analyze a recently proposed class of algorithms for the problem of sampling from probability distributions $μ^\ast$ in $\mathbb{R}^d$ with a Lebesgue density of the form $μ^\ast(x) \propto \exp(-f(Kx)-g(x))$, where $K$ is a linear operator and $f,g$ convex and non-smooth. The method is a generalization of the primal-dual hybrid gradient optimization algorithm to a sampling scheme. We give the iteration's continuous time limit, a stochastic differential equation in the joint primal-dual variable, and its mean field limit Fokker-Planck equation. Under mild conditions, the scheme converges to a unique stationary state in continuous and discrete time. Contrary to purely primal overdamped Langevin diffusion, the stationary state in continuous time does not have $μ^\ast$ as its primal marginal. Thus, further analysis is carried out to bound the bias induced by the partial dualization, and potentially correct for it in the diffusion. Time discretizations of the diffusion lead to implementable algorithms, but, as is typical in Langevin Monte Carlo methods, introduce further bias. We prove bounds for these discretization errors, which allow to give convergence results relating the produced samples to the target. We demonstrate our findings numerically first on small-scale examples in which we can exactly verify the theoretical results, and subsequently on typical examples of larger scale from Bayesian imaging inverse problems.
title Analysis of Primal-Dual Langevin Algorithms
topic Optimization and Control
35Q84, 47A52, 49N45, 60H10, 62F15, 65J22, 68U10
url https://arxiv.org/abs/2405.18098