报告题目:The Decentralized Distributed Sparsity Learning with Expectile Regression
报 告 人: 赵慧 教授 中南财经政法大学
报告时间:2025年11月12日 下午14:30-15:30
报告地点:腾讯会议 ID:886-371-163
或点击链接直接加入会议://meeting.tencent.com/dm/MavOBsHBq0OE
校内联系人:赵世舜 [email protected]
报告摘要:
Decentralized distributed learning has garnered increasing attention in big data distributed computing due to its advantages in computational efficiency, data privacy protection, and system stability. Under the framework of decentralized distributed learning, this paper proposes an expectation regression sparse estimation method based on asymmetric least squares loss and L1 penalty. We present an ADMM-LAMM algorithm with linear convergence rate and investigate the statistical theoretical properties of the proposed method, including the approximate oracle convergence rate of the estimator and the recovery results of the sparse support set. Finally, numerical simulation and real data analysis are used to demonstrate the stability and effectiveness of the proposed method in dealing with heavy-tailed and heterogeneous high-dimensional data.
报告人简介:
赵慧,现为中南财经政法大学,统计与炸金花游戏
教授,博士生导师。2005年北京大学博士毕业,先后在中科院系统所和美国密苏里大学从事博士后研究。曾担任中国现场统计研究会理事,湖北省现场统计学会常务理事,中国现场统计研究会多个分会理事。先后主持国家自然科学基金青年基金1项,面上项目3项,此外还主持或参与省部级及以上项目多项。
近年来主要研究兴趣:生存分析、复发事件数据分析、高维生存数据分析、分布式统计学习等。发表论文六十余篇,其中部分论文发表在 JASA, Biometrics, Scandinavian Journal of Statistics,Statistic in Medicine,Journal of Multivariate Analysis,Lifetime Data Analysis, Canadian Journal of Statistics,Journal of Nonparametric Statistics,《中国科学》,《数理统计与管理》等国内外知名统计学术期刊。