We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Data Structures and Algorithms

Title: Simple Parallel Algorithms for Single-Site Dynamics

Abstract: The single-site dynamics are a canonical class of Markov chains for sampling from high-dimensional probability distributions, e.g. the ones represented by graphical models.
We give a simple and generic parallel algorithm that can faithfully simulate single-site dynamics. When the chain asymptotically satisfies the $\ell_p$-Dobrushin's condition, specifically, when the Dobrushin's influence matrix has constantly bounded $\ell_p$-induced operator norm for an arbitrary $p\in[1,\infty]$, the parallel simulation of $N$ steps of single-site updates succeeds within $O\left({N}/{n}+\log n\right)$ depth of parallel computing using $\tilde{O}(m)$ processors, where $n$ is the number of sites and $m$ is the size of graphical model. Since the Dobrushin's condition is almost always satisfied asymptotically by mixing chains, this parallel simulation algorithm essentially transforms single-site dynamics with optimal $O(n\log n)$ mixing time to RNC algorithms for sampling. In particular we obtain RNC samplers, for the Ising models on general graphs in the uniqueness regime, and for satisfying solutions of CNF formulas in a local lemma regime. With non-adaptive simulated annealing, these RNC samplers can be transformed routinely to RNC algorithms for approximate counting.
A key step in our parallel simulation algorithm, is a so-called "universal coupling" procedure, which tries to simultaneously couple all distributions over the same sample space. We construct such a universal coupling, that for every pair of distributions the coupled probability is at least their Jaccard similarity. We also prove this is optimal in the worst case. The universal coupling and its applications are of independent interests.
Subjects: Data Structures and Algorithms (cs.DS)
Cite as: arXiv:2111.04044 [cs.DS]
  (or arXiv:2111.04044v1 [cs.DS] for this version)

Submission history

From: Yitong Yin [view email]
[v1] Sun, 7 Nov 2021 10:27:16 GMT (57kb,D)

Link back to: arXiv, form interface, contact.