Multiple Change Point Detection for Correlated  High-Dimensional Observations via the Largest Eigenvalue

主讲人：潘光明

讲座时间：2017年11月20日（周一）下午13:40-15:00  

讲座地点：浙江工商大学综合楼615会议室

主讲人简介：

潘光明，新加坡南洋理工大学副教授。2005年7月博士毕业于中国科学技术大学，数理统计专业；自2008年以来，在新加坡南洋理工大学工作，博士生导师，研究领域：高维统计推断、随机矩阵理论、计量经济学、多元统计、应用概率等。已在《Journal of the Royal Statistical Society Series  B》、《Journal of the American Statistical  Association》、《Annals of Statistics》、《Annals of  Probability》、《Annals of Applied Probability》、《IEEE  Transactions on Information Theory》、《IEEE  Transactions on Signal Processing 等SCI期刊杂志上发表50余篇专业学术论文。  

摘要：

We  propose to deal with a mean vector change point detection problem from a new  perspective via the largest eigenvalue when the data dimension p is comparable  to the sample size n. An optimization approach is proposed to figure out both  the unknown number of change points and multiple change point positions  simultaneously. Moreover, an adjustment term is introduced to handle sparse  signals when the change only appears in few components out of the p dimensions.  The computation time is controlled at $O(n^2)$ by adopting a dynamic  programming, regardless of the true number of change points $k_0$. Theoretical  results are developed and various simulations are conducted to show the  effectiveness of our method.