学院动态

学术报告信息发布

当前位置: 首页 >> 学术报告信息发布 >> 正文

Optimal Subsampling for Big Data: from ‘Static Data’ to ‘Data Streams’

发布日期:2024-10-11    作者:     点击:

报告题目:Optimal Subsampling for Big Data: from Static Datato Data Streams

报告时间:20241012上午1000

报告地点:南湖校区教学科研楼404

主办单位:数学与统计学院/科研处

报告人:艾明要

报告人简介:艾明要北京大学数学科学学院统计学二级教授、博士生导师。兼任全国应用统计专业学位研究生教育指导委员会委员、培养组组长,中国现场统计研究会副理事长,中国数学会概率统计学会第十一届理事会秘书长,中国统计学会常务理事。担任四个国际重要SCI期刊Stat SinicaJSPISPLStat编委,国内核心期刊 《系统科学与数学》、《数理统计与管理》、《数学进展》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理论与算法、试验设计与分析、计算机仿真试验与建模、应用统计的教学和研究工作,在AOSJASABiometrika、《中国科学》等国内外重要期刊发表学术论文八十余篇。主持国家自然科学基金重点项目1项、重点项目子课题1项、面上项目5项,参与完成科技部重点研发计划项目2项。北京大学通识教育核心课程主讲教师,两次获得北京大学优秀博士学位论文指导教师,获北京市高等学校优秀教学成果二等奖。

摘要:Subsampling methods are effective techniques to reduce computational burden and maintain statistical inference efficiency for big data. In this talk, we will review different subsampling techniques for efficiently dealing with different types of big data, not only for different inferential models from linear model, to generalized linear model, and to estimation equations, but also for different types of data from static data to data streams. To deal with the situation that the full data are stored in different blocks or at multiple locations, a distributed subsampling framework is developed, in which statistics are computed simultaneously on smaller partitions of the full data. Finally, the proposed strategies are illustrated and evaluated through numerical experiments on both simulated and real data sets.


上一条:Quantile-Matched DC in Massive Data Regression

下一条:On heavy-tailed matrix variate regression models and diagnostics