Optimal Distributed Subsampling Techniques for Big Data Analysis大数据分析的分布式抽样技术

24.01.2022  19:13
主  讲  人  : 艾明        教授

活动时间: 01月25日10时00分       

地            点  : 腾讯会议:175 844 885

讲座内容:

Subsampling methods are effective techniques to reducecomputational burden and maintain statistical inference efficiency for bigdata. In this talk, we will review different subsampling techniques fordifferent models from linear model, to generalized linear model, and toestimation equations. If the data volume is so large that nonuniformsubsampling probabilities cannot be calculated all at once, subsampling withreplacement is infeasible to implement. This problem is solved by using a new subsamplingwithout replacement, called Poisson subsampling. To deal with the situationthat the full data are stored in different blocks or at multiple locations, adistributed subsampling framework is developed, in which statistics arecomputed simultaneously on smaller partitions of the full data. Finally, theproposed strategies are illustrated and evaluated through numerical experimentson both simulated and real data sets.

主讲人介绍:

艾明,北京大学数学科学学院统计学教授、博士生导师,兼任全国应用统计专业硕士学位研究生教育指导委员会委员,中国现场统计研究会第十一届理事会副理事长,试验设计分会理事长,高维数据统计分会副理事长,中国数学会第十三届理事会理事,中国概率统计学会秘书长,中国数学会均匀设计分会副主任等。担任4个国际重要SCI期刊StatisticaSinica、JSPI、SPL和Stat的副主编,国内核心期刊 《系统科学与数学》、《数理统计与管理》编委,科学出版社《统计与数据科学丛书》编委。主要从事大数据采样技术、试验设计与分析、应用概率统计的教学和研究工作,在AoS、JASA、Biometrika、《中国科学》等国内外重要期刊发表学术论文七十余篇。主持国家自然科学基金重点项目1项,主持国家自然科学基金面上项目6项,参与完成科技部重点研发计划(973)项目2项。

发布时间:2022-01-24 16:57:56