学科科研

【学术讲座】Some Model-free Subsampling Methods for Massive Data

发布日期:2025-09-22 16:03:38   来源:统计与数学学院   点击量:


报告主题:Some Model-free Subsampling Methods for Massive Data

时间:202592415:00-16:00

地点:腾讯会议:586-677-536

报告人周永道

 

报告内容简介:

    High-quality training data is crucial for the effectiveness of big data analytics algorithms. Effective data collection methods can be used to extract high-quality subsamples from massive data, reducing training costs and enabling rapid model training. Ture models are often nonlinear, then model-free subsampling methods are needed. Experimental design is an important type of data collection method. This report will introduce some robust experimental design methods and propose some model-free subsampling methods under similarity and data shift. Simulations and practical examples demonstrate that the resulting subsamples have good performance.

 

主讲人简介:

        周永道,男,南开大学统计与数据科学学院教授、博导,统计学系主任,入选国家高水平人才青年项目、天津市创新类领军人才、南开大学百青。研究方向为试验设计和大数据分析。主持过 5 项国家自然科学基金、项天津市自然科学基金重点项目和其它 10 余项纵横向项目。曾访问 UCLA  5 所境外高校。在统计学和机器学习顶级期刊JRSSBJASABiometrikaJMLRTKDETNNLS 及中国科学等国内外期刊发表学术论文80余篇;合作出版了 9 部中英文专著和教材。曾获全国统计科学研究优秀成果奖一等奖、全国统计科学技术进步奖三等奖、天津市教学成果奖特等奖、华为火花奖。现为天津市现场统计研究会理事长,中国数学会理事、均匀设计分会副理事长。

分享到: