Top Views

Model-free feature screening for ultrahigh dimensional classification-Forum of Distinguished Scholars of Faculty of Science

Welcome the 60th Anniversary of the Founding of BJUT

Forum of Distinguished Scholars of Faculty of Science

Model-free feature screening for ultrahigh dimensional classification

Guest Speaker:Wang Qihua(from China Academy of Science)

Date:June 28,2020(Sunday)

Time:20:00-21:00

ID: Tencent meeting 158 789 349

Contract:In this lecture, a new model-free feature screening method based on classification accuracy of marginal classifiers is proposed for ultrahigh dimensional classification. Different from existing methods, which use the differences of means or differences of conditional cumulative distribution functions between classes as the screening indexes, we propose a new feature screening method to rank the importance of predictors based on classification accuracy of marginal classifiers. For each variable, we construct the corresponding marginal classifier according to the Bayes rule and thus classification accuracy of these marginal classifiers can be used as effective feature screening indexes to select all important variables. Not only for a fixed number of classes but also for a diverging number of classes, we can prove that the proposed method enjoys the sure screening property under some regularity conditions. Finally, simulations and the real data analysis well demonstrate good performance of the proposed method in comparison with existing methods.

Resume of Speaker:Wang Qihua is the Core Backbone Visiting Professor of Chinese Academy of Sciences, Ph.D. supervisor, winner of the National Science Fund for Distinguished Young Scholars, Distinguished Professor of Talent Scholar Program of Ministry of Education, candidate of Hundred Talents Program of Chinese Academy of Sciences, and one of first winners of the 100th Excellent China Doctoral Dissertations. He was on the faculty at Peking University and the University of Hong Kong. He has visited Carleton University in Canada, University of California, Davis, California State University, Los Angeles, Yale University, Washington University and Northwestern University in the United States, Humboldt University in Germany, Australian National University and the University of Sydney in Australia. He principally engaged in survival analysis, missing data analysis, statistical analysis of high-dimensional data and semi- or non-parametric statistical inference. He has published three books and more than 100 papers on international journals, including The Annals of Statistics, JASA and Biometrika. He serves as the president of High Dimensional Statistics Branch, the vice president of Survival Analysis and Biostatistics Branch, a member of IMS-China and IBS-China Committee successively, and he is also an editorial board member of some international and domestic academic journals.

Contact Person:Wang Teng Email:wangteng@bjut.edu.cn

Welcome to attend!