报告时间：2024年3月28日(周四)14:00-15:00
报告地点：中国人民大学明德主楼1016
报告主题：Variational Bayes approaches to model selection and estimation
报告摘要：
Model selection and estimation are some of the most fundamental tasks in analysis. The approaches to selection in the linear regression model context have been extensively developed since the 1970s and are currently widely in use. However, the ability of existing methods to produce efficient and effective analysis for high-dimensional data is still limited. Moreover, most of the current model selection techniques are developed for the linear regression model only, and thus may not be appropriate in other model contexts. In this talk, we present fast alternative methods for model selection and estimation based on the variational Bayes method. The proposed methods include mean field variational Bayes, fixed form variational Bayes, and collapsed variational Bayes. These methods possess desirable consistency properties under mild regularity conditions for estimators, while achieving efficient computation times comparable to popular variable selection approaches.
报告人简介：
尤翀，北京大学北京国际数学研究中心特聘副研究员、北京大学国家药品医疗器械监管科学研究院独立PI。主要研究方向为变分贝叶斯，混合模型，变量选择等，近年来也在传染病模型上投入大量工作，文章发表在Science Advances, JASA, Biometrics等杂志。
