您所在的位置:首页 > 中心动态 > 信息通知

【信息通知】哥伦比亚大学冯阳副教授学术讲座

2019-7-18 10:23:16

 

为了进一步营造科学研究范围,开拓国际学术视野,提升研究能力、学术水平和国际交流能力,大连海事大学航运经济与管理学院、综合交通运输协同创新中心于2019720日邀请了哥伦比亚大学冯阳副教授来我校做学术讲座。

讲座嘉宾:冯阳,现就职于美国哥伦比亚大学统计系,任副教授

讲座时间:2019720 10001130

讲座地点:大连海事大学大学生活动中心610

讲座主题:Model Averaging Estimation for Nonlinear Regression Models

专家简介:冯阳,现就职于美国哥伦比亚大学统计系,任副教授,美国统计协会(ASA)会员、数理统计协会(IMS)成员、国际统计协会成员、国际中国统计协会(ICSA)终身会员。2016年获NSF CAREER奖,2015年获Lenfest青年教授发展奖,2012年获新世界数学银奖。本科毕业于中国科学技术大学少年班, 2010年获得普林斯顿大学运筹与金融工程系理学博士学位,师从国际著名统计学家范剑青教授。研究领域主要包括高维统计学习,网络模型,非参数、半参数方法以及生物信息学等。

任《Journal of Business and Economic Statistics》、《Statistica Sinica》、 Computational Statistics and Data Analysis》、《Statistical Analysis and Data Mining》副主编。2018年发表《A Kronecker Product Model for Repeated Pattern Detection on 2D Urban Images》、《Neyman-Pearson classification algorithms and NP receiver operating characteristics.》等研究成果。2017年发表《Model Selection for High Dimensional Quadratic Regression via Regularization.》、《Regularization after retention in ultrahigh dimensional linear regression models.》等。2016年发表《Feature Augmentation via Nonparametrics and Selection (FANS) in High Dimensional Classification.》、《Neyman-Pearson Classification under High-Dimensional Settings》等。所编写软件包在CRAN上被下载超过45,000次。

AbstractThis paper considers the problem of model averaging estimation for regression models that are possibly nonlinear in their parameters and variables. We develop a nonlinear model averaging method (NMA) and propose a new weight-choosing criterion named NIC_MA. We show that up to a constant term, NIC_MA is an asymptotically unbiased estimator of the risk function under nonlinear settings. We also prove the optimality of NMA in terms of weight selection in the scenarios of both fixed and increasing number of models and obtain the asymptotic distribution of the out-of-sample forecast. Monte Carlo experiments show that NMA leads to relatively lower risks compared with state-of-the-art model selection and model averaging methods across various settings. Finally, the NMA method is applied to predict the individual wage and is shown to have the lowest prediction errors.