哈尔滨工业大学（深圳）学术讲座

演讲人Speaker： 彭衡    

题目Title: Automatic Change Point Detection and Segment

Estimation via Variational Bayesian Model Selection

时间Date：  2023 年 11 月 8 日       Time：16：00-17：00

地点Venue：  H 栋 307 室

内容摘要Abstract： 

Change-point detection has long been an active research areas, especially in the Big Data era, where data streams are usually non-stationary. However, many existing methods require a preset number of change point or using certain stopping threshold, therefore limited to some model selection criterion. In this paper, we introduce an offline Bayesian change point model and associated scalable variational EM algorithm which can automatically estimate the number of change points and within parameters in each segment. The comprehensive simulations for a normal mean-variance shift model and a discrete Poisson model and a real application in finance demonstrate advantages of our approach. All the change-point locations and posterior inferences indicate the proposed method is comparable in location and parameter estimation, and computational efficiency over existing methods.

个人简介(About the speaker):

彭衡，香港浸会大学数学系教授，博士生导师。2003年从香港中文大学获得统计学博士学位，2003年-2006年在普林斯顿大学做博士后。他主要从事非参数与半参数模型、模型选择、高维数据建模、混合模型等领域的研究。他是IMS的会员，2011-2014担任Statistica Sinica副主编，现为Computational Statistics and Data Analysis副主编。曾做过Annals of Statistics、JASA、JRSSB、Biometrika、Statistica Sinica等国际期刊的评审。在统计学国际顶级期刊Annals of Statistics、JASA、Biometrika、Statistica Sinica、TEST 和Computational Statistics and Data Analysis等期刊上发表学术论文。