财大数统学院
下载中心
讲座 您现在所在的位置: 首页资源分类讲座

吕绍高教授讲座公告
来源:中南财经政法大学统计与数学学院    编辑:佚名    时间:2017-4-25    点击数:

报 告 人:吕绍高(西南财经大学)

报告时间:4月29号上午10:00—11:00

报告地点:文波楼304统数学院实验室

报告题目:kernel-based feature selection approach for classification without model specification

摘    要:This talk presents a feature selection approach for classification allowing for complex data structures. The proposed approach consists of a regularized learning algorithm with two-fold penalties without any model assumption. Then, our learning scheme is equivalent to a convex finite dimensional optimization, even if it is defined directly in a general reproducing kernel Hilbert space. Furthermore, we develop an efficient numerical algorithm for solving estimated coefficients, which is verified well in both simulated examples and real data compared to several existing methods. Finally, we prove estimation consistency and prediction ability of our proposed estimator.


    主讲人简介:

   吕绍高,西南财经大学统计学院教授,博士生导师。2011年获得中国科技大学-香港城市大学联合培养博士学位。研究兴趣主要是统计机器学习与数据挖掘,已经在统计和机器学习类国际顶级杂志《Annals of statistics》,《Journal of machine learning research》等发表论文近20篇。

 

上一篇:讲座通知
下一篇:讲座预告