We show how to sample in parallel from a distribution $π$ over $\mathbb R^d$ that satisfies a log-Sobolev inequality and has a smooth log-density, by parallelizing the Langevin (resp. underdamped Langevin) algorithms. We show that our algorithm outputs samples from a distribution $\hatπ$ that is close to $π$ in Kullback--Leibler (KL) divergence (resp. total variation (TV) distance), while using only $\log(d)^{O(1)}$ parallel rounds and $\widetilde{O}(d)$ (resp. $\widetilde O(\sqrt d)$) gradient evaluations in total. This constitutes the first parallel sampling algorithms with TV distance guarantees.
For our main application, we show how to combine the TV distance guarantees of our algorithms with prior works and obtain RNC sampling-to-counting reductions for families of discrete distribution on the hypercube $\{\pm 1\}^n$ that are closed under exponential tilts and have bounded covariance. Consequently, we obtain an RNC sampler for directed Eulerian tours and asymmetric determinantal point processes, resolving open questions raised in prior works.
Logconcave functions represent the current frontier of efficient algorithms for sampling, optimization and integration in R^n. Efficient sampling algorithms to sample according to a probability density (to which the other two problems can be reduced) relies on good isoperimetry which is known to hold for arbitrary logconcave densities. In this paper, we extend this frontier in two ways: first, we characterize convexity-like conditions that imply good isoperimetry, i.e., what condition on function values along every line guarantees good isoperimetry? The answer turns out to be the set of (1/(n-…
We study the Proximal Langevin Algorithm (PLA) for sampling from a probability distribution $ν= e^{-f}$ on $\mathbb{R}^n$ under isoperimetry. We prove a convergence guarantee for PLA in Kullback-Leibler (KL) divergence when $ν$ satisfies log-Sobolev inequality (LSI) and $f$ has bounded second and third derivatives. This improves on the result for the Unadjusted Langevin Algorithm (ULA), and matches the fastest known rate for sampling under LSI (without Metropolis filter) with a better dependence on the LSI constant. We also prove convergence guarantees for PLA in Rényi divergence of order $q >…
We study distributed protocols for finding all pairs of similar vectors in a large dataset. Our results pertain to a variety of discrete metrics, and we give concrete instantiations for Hamming distance. In particular, we give improved upper bounds on the overhead required for similarity defined by Hamming distance $r>1$ and prove a lower bound showing qualitative optimality of the overhead required for similarity over any Hamming distance $r$. Our main conceptual contribution is a connection between similarity search algorithms and certain graph-theoretic quantities. For our upper bounds, we …
Georgia Institute of Technology（佐治亚理工学院）；Yale University（耶鲁大学）
We study the Unadjusted Langevin Algorithm (ULA) for sampling from a probability distribution $ν= e^{-f}$ on $\mathbb{R}^n$. We prove a convergence guarantee in Kullback-Leibler (KL) divergence assuming $ν$ satisfies a log-Sobolev inequality and the Hessian of $f$ is bounded. Notably, we do not assume convexity or bounds on higher derivatives. We also prove convergence guarantees in Rényi divergence of order $q > 1$ assuming the limit of ULA satisfies either the log-Sobolev or Poincaré inequality. We also prove a bound on the bias of the limiting distribution of ULA assuming third-order smooth…
We present randomized algorithms for sampling the standard Gaussian distribution restricted to a convex set and for estimating the Gaussian measure of a convex set, in the general membership oracle model. The complexity of integration is $O^*(n^3)$ while the complexity of sampling is $O^*(n^3)$ for the first sample and $O^*(n^2)$ for every subsequent sample. These bounds improve on the corresponding state-of-the-art by a factor of $n$. Our improvement comes from several aspects: better isoperimetry, smoother annealing, avoiding transformation to isotropic position and the use of the "speedy wa…
This paper investigates parallel random sampling from a potentially-unending data stream whose elements are revealed in a series of element sequences (minibatches). While sampling from a stream was extensively studied sequentially, not much has been explored in the parallel context, with prior parallel random-sampling algorithms focusing on the static batch model. We present parallel algorithms for minibatch-stream sampling in two settings: (1) sliding window, which draws samples from a prespecified number of most-recently observed elements, and (2) infinite window, which draws samples from al…
A deterministic BSP algorithm for constructing the suffix array of a given string is presented, based on a technique which we call accelerated sampling. It runs in optimal O(n/p) local computation and communication, and requires a near optimal O(log log p) synchronisation steps. The algorithm provides an improvement over the synchronisation costs of existing algorithms, and reinforces the importance of the sampling technique.
We study the problem of sampling a uniformly random directed rooted spanning tree, also known as an arborescence, from a possibly weighted directed graph. Classically, this problem has long been known to be polynomial-time solvable; the exact number of arborescences can be computed by a determinant [Tut48], and sampling can be reduced to counting [JVV86, JS96]. However, the classic reduction from sampling to counting seems to be inherently sequential. This raises the question of designing efficient parallel algorithms for sampling. We show that sampling arborescences can be done in RNC. For se…
We consider the problem of sampling $n$ numbers from the range $\{1,\ldots,N\}$ without replacement on modern architectures. The main result is a simple divide-and-conquer scheme that makes sequential algorithms more cache efficient and leads to a parallel algorithm running in expected time $\mathcal{O}(n/p+\log p)$ on $p$ processors, i.e., scales to massively parallel machines even for moderate values of $n$. The amount of communication between the processors is very small (at most $\mathcal{O}(\log p)$) and independent of the sample size. We also discuss modifications needed for load balanci…
Data structures for efficient sampling from a set of weighted items are an important building block of many applications. However, few parallel solutions are known. We close many of these gaps both for shared-memory and distributed-memory machines. We give efficient, fast, and practicable parallel algorithms for building data structures that support sampling single items (alias tables, compressed data structures). This also yields a simplified and more space-efficient sequential algorithm for alias table construction. Our approaches to sampling $k$ out of $n$ items with/without replacement and…
Motivated by a sampling problem basic to computational statistical inference, we develop a nearly optimal algorithm for a fundamental problem in spectral graph theory and numerical analysis. Given an $n\times n$ SDDM matrix ${\bf \mathbf{M}}$, and a constant $-1 \leq p \leq 1$, our algorithm gives efficient access to a sparse $n\times n$ linear operator $\tilde{\mathbf{C}}$ such that $${\mathbf{M}}^{p} \approx \tilde{\mathbf{C}} \tilde{\mathbf{C}}^\top.$$ The solution is based on factoring ${\bf \mathbf{M}}$ into a product of simple and sparse matrices using squaring and spectral sparsificatio…
The uniform sampling of simple graphs matching a prescribed degree sequence is an important tool in network science, e.g. to construct graph generators or null-models. Here, the Edge Switching Markov Chain (ES-MC) is a common choice. Given an arbitrary simple graph with the required degree sequence, ES-MC carries out a large number of small changes, called edge switches, to eventually obtain a uniform sample. In practice, reasonably short runs efficiently yield approximate uniform samples. In this work, we study the problem of executing edge switches in parallel. We discuss parallelizations of…
\emph{Sampling} constitutes an important tool in a variety of areas: from machine learning and combinatorial optimization to computational physics and biology. A central class of sampling algorithms is the \emph{Markov Chain Monte Carlo} method, based on the construction of a Markov chain with the desired sampling distribution as its stationary distribution. Many of the traditional Markov chains, such as the \emph{Glauber dynamics}, do not scale well with increasing dimension. To address this shortcoming, we propose a simple local update rule based on the Glauber dynamics that leads to efficie…
The k-center problem is one of several classic NP-hard clustering questions. For contemporary massive data sets, RAM-based algorithms become impractical. And although there exist good sequential algorithms for k-center, they are not easily parallelizable. In this paper, we design and implement parallel approximation algorithms for this problem. We observe that Gonzalez's greedy algorithm can be efficiently parallelized in several MapReduce rounds; in practice, we find that two rounds are sufficient, leading to a 4-approximation. We contrast this with an existing parallel algorithm for k-center…
We propose new sequential sorting operations by adapting techniques and methods used for designing parallel sorting algorithms. Although the norm is to parallelize a sequential algorithm to improve performance, we adapt a contrarian approach: we employ parallel computing techniques to speed up sequential sorting. Our methods can also work for multi-core sorting with minor adjustments that do not necessarily require full parallelization of the original sequential algorithm. The proposed approach leads to the development of asymptotically efficient deterministic and randomized sorting operations…
An alias table is a data structure that allows for efficiently drawing weighted random samples in constant time and can be constructed in linear time. The PSA algorithm by Hübschle-Schneider and Sanders is able to construct alias tables in parallel on the CPU. In this report, we transfer the PSA algorithm to the GPU. Our construction algorithm achieves a speedup of 17 on a consumer GPU in comparison to the PSA method on a 16-core high-end desktop CPU. For sampling, we achieve an up to 24 times higher throughput. Both operations also require several times less energy than on the CPU. Adaptation…
We examine the problem of almost-uniform sampling proper $q$-colorings of a graph whose maximum degree is $Δ$. A famous result, discovered independently by Jerrum(1995) and Salas and Sokal(1997), is that, assuming $q > (2+δ) Δ$, the Glauber dynamics (a.k.a. single-site dynamics) for this problem has mixing time $O(n \log n)$, where $n$ is the number of vertices, and thus provides a nearly linear time sampling algorithm for this problem. A natural question is the extent to which this algorithm can be parallelized. Previous work Feng, Sun and Yin [PODC'17] has shown that a $O(Δ\log n)$ time para…
Betweenness centrality is one of the most popular vertex centrality measures in network analysis. Hence, many (sequential and parallel) algorithms to compute or approximate betweenness have been devised. Recent algorithmic advances have made it possible to approximate betweenness very efficiently on shared-memory architectures. Yet, the best shared-memory algorithms can still take hours of running time for large graphs, especially for graphs with a high diameter or when a small relative error is required. In this work, we present an MPI-based generalization of the state-of-the-art shared-memor…
