arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2308.09769stat.COcs.DC
收藏

Pigeons.jl:从难处理分布中进行分布式采样

Pigeons.jl: Distributed Sampling From Intractable Distributions

Nikola Surjanovic, Miguel Biron-Lattes, Paul Tiede, Saifuddin Syed, Trevor Campbell, Alexandre Bouchard-Côté

AI总结:

本文提出 Julia 软件包 Pigeons.jl,通过单线程、多线程及 MPI 分布式计算从复杂难处理概率分布中采样,并保证给定种子下的强并行不变性以支持可复现性与验证。

摘要:

我们介绍了一个名为 Pigeons.jl 的软件包,它提供了一种利用分布式计算从复杂概率分布中获取样本的方法,例如贝叶斯推断中出现的多峰后验分布以及统计力学中的高维分布。Pigeons.jl 提供了简单的 API,可在单线程、多线程和/或由数千台通过 MPI 通信的机器组成的分布式环境中执行此类计算。此外,Pigeons.jl 保证了一项我们称为强并行不变性的性质:对于给定种子,无论线程和进程数量如何,输出都完全相同,这对科学可复现性和软件验证至关重要。我们描述了 Pigeons.jl 的关键特性,以及为实现满足强并行不变性的分布式随机算法所采用的方法。

英文摘要:

We introduce a software package, Pigeons.jl, that provides a way to leverage distributed computation to obtain samples from complicated probability distributions, such as multimodal posteriors arising in Bayesian inference and high-dimensional distributions in statistical mechanics. Pigeons.jl provides simple APIs to perform such computations single-threaded, multi-threaded, and/or distributed over thousands of MPI-communicating machines. In addition, Pigeons.jl guarantees a property that we call strong parallelism invariance: the output for a given seed is identical irrespective of the number of threads and processes, which is crucial for scientific reproducibility and software validation. We describe the key features of Pigeons.jl and the approach taken to implement a distributed and randomized algorithm that satisfies strong parallelism invariance.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑