PreprintCC BY 4.0arXiv
Jensen-Bregman split Gibbs sampling
Abstract
Split Gibbs sampling (SGS) is a versatile tool to sample from complex posterior distributions, such as those arising in imaging inverse problems. However, existing SGS instances are limited to quadratic variable couplings, restricting their applicability to Gaussian likelihoods. This letter genearalizes SGS to non-Euclidean geometry by introducing a coupling kernel defined through a Jensen-Bregman divergence. This framework not only enables posterior sampling in these settings but also unifies several \emph{ad hoc} sampling schemes from the literature under a principled theoretical foundation.
Related
- Learning to Replace MCMC in Split-Gibbs Diffusion Posterior Sampling via Deep Unfolding
- Kinetic Langevin Meets Split Gibbs: Accelerated Posterior Sampling for Imaging Inverse Problems with Diffusion Priors
- Non-Uniform Random Scans in Gibbs Sampling and CAVI
- On skew-symmetric distributions and their use in Monte Carlo sampling algorithms: coordinate-free, Gibbs-style and manifold versions of the Barker proposal
- Gibbs Sampling for Bayesian Generalized Poisson Matrix Factorization
BibTeX
@misc{dobigeon2026jensenbregman,
title = {{Jensen-Bregman split Gibbs sampling}},
author = {Nicolas Dobigeon and Mame Diarra Fall and Eric Barat},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.15853},
eprint = {2610.15853},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.15853},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Jensen-Bregman split Gibbs sampling
AU - Nicolas Dobigeon
AU - Mame Diarra Fall
AU - Eric Barat
PY - 2026
DA - 2026/10/11
PB - arXiv
DO - 10.48550/arxiv.2610.15853
UR - https://arxiv.org/abs/2610.15853
AB - Split Gibbs sampling (SGS) is a versatile tool to sample from complex posterior distributions, such as those arising in imaging inverse problems. However, existing SGS instances are limited to quadratic variable couplings, restricting their applicability to Gaussian likelihoods. This letter genearalizes SGS to non-Euclidean geometry by introducing a coupling kernel defined through a Jensen-Bregman divergence. This framework not only enables posterior sampling in these settings but also unifies several \emph{ad hoc} sampling schemes from the literature under a principled theoretical foundation.
ER -