吉布斯采样中的扫描顺序：扫描顺序起作用的模型及其影响程度的界限
Scan Order in Gibbs Sampling: Models in Which it Matters and Bounds on How Much.
作者信息
He Bryan, De Sa Christopher, Mitliagkas Ioannis, Ré Christopher
机构信息
Stanford University.
出版信息
Adv Neural Inf Process Syst. 2016;29.
Abstract
Gibbs sampling is a Markov Chain Monte Carlo sampling technique that iteratively samples variables from their conditional distributions. There are two common scan orders for the variables: random scan and systematic scan. Due to the benefits of locality in hardware, systematic scan is commonly used, even though most statistical guarantees are only for random scan. While it has been conjectured that the mixing times of random scan and systematic scan do not differ by more than a logarithmic factor, we show by counterexample that this is not the case, and we prove that that the mixing times do not differ by more than a polynomial factor under mild conditions. To prove these relative bounds, we introduce a method of augmenting the state space to study systematic scan using conductance.
摘要
吉布斯采样是一种马尔可夫链蒙特卡罗采样技术，它从变量的条件分布中迭代地进行采样。变量有两种常见的扫描顺序：随机扫描和系统扫描。由于硬件中局部性的优势，即使大多数统计保证仅适用于随机扫描，系统扫描仍被普遍使用。虽然有人推测随机扫描和系统扫描的混合时间相差不超过对数因子，但我们通过反例表明情况并非如此，并且我们证明在温和条件下混合时间相差不超过多项式因子。为了证明这些相对界限，我们引入了一种扩充状态空间的方法，以使用电导来研究系统扫描。
相似文献
1
Scan Order in Gibbs Sampling: Models in Which it Matters and Bounds on How Much.吉布斯采样中的扫描顺序：扫描顺序起作用的模型及其影响程度的界限
Adv Neural Inf Process Syst. 2016;29.
2
Replica exchange and expanded ensemble simulations as Gibbs sampling: simple improvements for enhanced mixing.复制交换和扩展系综模拟作为吉布斯抽样：增强混合的简单改进。
J Chem Phys. 2011 Nov 21;135(19):194110. doi: 10.1063/1.3660669.
3
Rapidly Mixing Gibbs Sampling for a Class of Factor Graphs Using Hierarchy Width.基于层次宽度的一类因子图的快速混合吉布斯采样
Adv Neural Inf Process Syst. 2015 Dec;28:3079-3087.
4
Ensuring Rapid Mixing and Low Bias for Asynchronous Gibbs Sampling.确保异步吉布斯采样的快速混合和低偏差。
JMLR Workshop Conf Proc. 2016;48:1567-1576.
5
Gibbs ensembles for incompatible dependency networks.用于不相容依赖网络的吉布斯系综。
Wiley Interdiscip Rev Comput Stat. 2013 Nov-Dec;5(6):478-485. doi: 10.1002/wics.1273. Epub 2013 Aug 31.
6
Geometric ergodicity of a hybrid sampler for Bayesian inference of phylogenetic branch lengths.用于系统发育分支长度贝叶斯推断的混合采样器的几何遍历性
Math Biosci. 2015 Oct;268:9-21. doi: 10.1016/j.mbs.2015.07.002. Epub 2015 Aug 7.
7
Rapid Adiabatic Preparation of Injective Projected Entangled Pair States and Gibbs States.快速绝热制备注射纠缠对态和吉布斯态。
Phys Rev Lett. 2016 Feb 26;116(8):080503. doi: 10.1103/PhysRevLett.116.080503. Epub 2016 Feb 25.
8
Markov chain Monte Carlo sampling of gene genealogies conditional on unphased SNP genotype data.基于未分型单核苷酸多态性（SNP）基因型数据的基因谱系的马尔可夫链蒙特卡罗抽样。
Stat Appl Genet Mol Biol. 2013 Oct 1;12(5):559-81. doi: 10.1515/sagmb-2012-0011.
9
Marathon: An Open Source Software Library for the Analysis of Markov-Chain Monte Carlo Algorithms.《马拉松：用于马尔可夫链蒙特卡罗算法分析的开源软件库》
PLoS One. 2016 Jan 29;11(1):e0147935. doi: 10.1371/journal.pone.0147935. eCollection 2016.
10
A Monte Carlo Metropolis-Hastings algorithm for sampling from distributions with intractable normalizing constants.一种用于从具有难以处理的归一化常数的分布中进行抽样的蒙特卡罗 metropolis-hastings 算法。
Neural Comput. 2013 Aug;25(8):2199-234. doi: 10.1162/NECO_a_00466. Epub 2013 Apr 22.
引用本文的文献
1
Finding signatures of low-dimensional geometric landscapes in high-dimensional cell fate transitions.在高维细胞命运转变中寻找低维几何景观的特征。
bioRxiv. 2025 Jun 4:2025.05.27.656406. doi: 10.1101/2025.05.27.656406.
2
Hot Topic Recognition of Health Rumors Based on Anti-Rumor Articles on the WeChat Official Account Platform: Topic Modeling.基于微信公众号反谣言文章的健康谣言热点话题识别：主题建模。
J Med Internet Res. 2023 Sep 21;25:e45019. doi: 10.2196/45019.
3
netANOVA: novel graph clustering technique with significance assessment via hierarchical ANOVA.netANOVA：一种新颖的图聚类技术，通过层次 ANOVA 进行显著性评估。
Brief Bioinform. 2023 Mar 19;24(2). doi: 10.1093/bib/bbad029.
4
Health Communication through Chinese Media on E-Cigarette: A Topic Modeling Approach.通过中文媒体进行的电子烟健康传播：一种主题建模方法。
Int J Environ Res Public Health. 2022 Jun 21;19(13):7591. doi: 10.3390/ijerph19137591.
5
Accelerating All-Atom Simulations and Gaining Mechanistic Understanding of Biophysical Systems through State Predictive Information Bottleneck.通过状态预测信息瓶颈加速全原子模拟并获得生物物理系统的机制理解。
J Chem Theory Comput. 2022 May 10;18(5):3231-3238. doi: 10.1021/acs.jctc.2c00058. Epub 2022 Apr 6.
6
Health Communication About Hospice Care in Chinese Media: Digital Topic Modeling Study.中文媒体中的临终关怀健康传播：数字主题建模研究。
JMIR Public Health Surveill. 2021 Oct 21;7(10):e29375. doi: 10.2196/29375.
7
Health Communication Through News Media During the Early Stage of the COVID-19 Outbreak in China: Digital Topic Modeling Approach.中国新冠疫情初期通过新闻媒体进行的健康传播：数字主题建模方法
J Med Internet Res. 2020 Apr 28;22(4):e19118. doi: 10.2196/19118.
8
Data Analysis and Visualization of Newspaper Articles on Thirdhand Smoke: A Topic Modeling Approach.报纸上关于三手烟文章的数据分析与可视化：一种主题建模方法
JMIR Med Inform. 2019 Jan 29;7(1):e12414. doi: 10.2196/12414.
本文引用的文献
1
Rapidly Mixing Gibbs Sampling for a Class of Factor Graphs Using Hierarchy Width.基于层次宽度的一类因子图的快速混合吉布斯采样
Adv Neural Inf Process Syst. 2015 Dec;28:3079-3087.
2
Stochastic relaxation, gibbs distributions, and the bayesian restoration of images.随机松弛，吉布斯分布，以及贝叶斯图像恢复。
IEEE Trans Pattern Anal Mach Intell. 1984 Jun;6(6):721-41. doi: 10.1109/tpami.1984.4767596.
3
The BUGS project: Evolution, critique and future directions.BUGS 项目：演化、批判与未来方向。
Stat Med. 2009 Nov 10;28(25):3049-67. doi: 10.1002/sim.3680.
4
Finding scientific topics.寻找科学主题。
Proc Natl Acad Sci U S A. 2004 Apr 6;101 Suppl 1(Suppl 1):5228-35. doi: 10.1073/pnas.0307752101. Epub 2004 Feb 10.
5
Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm.通过隐马尔可夫随机场模型和期望最大化算法对脑部磁共振图像进行分割。
IEEE Trans Med Imaging. 2001 Jan;20(1):45-57. doi: 10.1109/42.906424.
Zotero 插件