Nesterov smoothing for sampling without smoothness

08/15/2022
by   Jiaojiao Fan, et al.
0

We study the problem of sampling from a target distribution in ℝ^d whose potential is not smooth. Compared with the sampling problem with smooth potentials, this problem is more challenging and much less well-understood due to the lack of smoothness. In this paper, we propose a novel sampling algorithm for a class of non-smooth potentials by first approximating them by smooth potentials using a technique that is akin to Nesterov smoothing. We then utilize sampling algorithms on the smooth potentials to generate approximate samples from the original non-smooth potentials. With a properly chosen smoothing intensity, the accuracy of the algorithm is guaranteed. For strongly log-concave distributions, our algorithm can achieve ℰ error in Wasserstein-2 distance with complexity 𝒪( d^1/3/ℰ^5/3) . For log-concave distributions, we achieve ℰ error in total variation with complexity 𝒪( M_π d /ℰ^3) in expectation with M_π being the second moment of the target distribution. For target distributions satisfying the logarithmic-Sobolev inequality, our algorithm has complexity 𝒪( d /ℰ).

READ FULL TEXT

page 1

page 2

page 3

page 4

research
05/20/2022

A Proximal Algorithm for Sampling from Non-convex Potentials

We study sampling problems associated with non-convex potentials that me...
research
06/20/2019

Bounding the error of discretized Langevin algorithms for non-strongly log-concave targets

In this paper, we provide non-asymptotic upper bounds on the error of sa...
research
06/03/2021

Stein's method, smoothing and functional approximation

Stein's method for Gaussian process approximation can be used to bound t...
research
02/20/2023

Improved dimension dependence of a proximal algorithm for sampling

We propose a sampling algorithm that achieves superior complexity bounds...
research
05/29/2021

Rejection sampling from shape-constrained distributions in sublinear time

We consider the task of generating exact samples from a target distribut...
research
06/16/2020

Primal Dual Interpretation of the Proximal Stochastic Gradient Langevin Algorithm

We consider the task of sampling with respect to a log concave probabili...
research
03/22/2023

Non-asymptotic analysis of Langevin-type Monte Carlo algorithms

We study Langevin-type algorithms for sampling from Gibbs distributions ...

Please sign up or login with your details

Forgot password? Click here to reset