Zeroth-Order Non-Log-Concave Sampling with Variance Reduction and Applications to Inverse Problems
This work provides the first non-asymptotic convergence guarantees for zeroth-order non-log-concave sampling, which is significant for researchers and practitioners working on black-box inverse problems with pre-trained score-based generative priors.
This paper addresses the challenge of sampling from high-dimensional, non-log-concave distributions in black-box settings where gradient information is unavailable. The authors propose a variance-reduced zeroth-order Langevin sampling method that improves upon classical estimators by reducing variance and eliminating unfavorable dimensional dependence in batch size, leading to practical and stable sampling.
Sampling from high-dimensional, non-log-concave distributions with unnormalized densities remains a fundamental challenge in machine learning, particularly in black-box settings where gradient information is inaccessible or computationally prohibitive. While Langevin dynamics provides a principled framework for sampling when gradients are accessible, its extension to the black-box settings suffers from high variance and lacks non-asymptotic convergence guarantees for non-log-concave sampling. To address these limitations, we propose a variance-reduced zeroth-order Langevin sampling method. Our method employs a gradient estimator that substantially reduces the variance of the classical batched zeroth-order estimator and eliminates the unfavorable dimensional dependence of the batch size required for accurate estimation, enabling practical and stable sampling. We establish the first non-asymptotic convergence guarantees for zeroth-order non-log-concave sampling in terms of $\varepsilon$-relative Fisher information, and, under a Poincaré inequality assumption, squared total variation distance. We further propose ZO-APMC, a posterior sampling algorithm for black-box inverse problems with pre-trained score-based generative priors, establishing the first non-asymptotic convergence guarantees for such methods. We validate our theory through synthetic experiments and demonstrate strong empirical performance on practical linear and nonlinear inverse problems.