主 题：TensorRegression and Neuroimaging Analysis
主讲人：Associate Prof.Lexin Li
主持人：林华珍教授
时 间：2015年7月9日下午3点-4点
地 点：通博楼B212学术会议室
主办单位：统计研究中心 统计学院 科研处
主讲人简介：
Lexin Liobtained B.E. in Electrical Engineering at Zhejiang University, P.R.China, in1998, and Ph.D. in Statistics at School of Statistics, University of Minnesota,in 2003. He then worked as a Postdoctoral Researcher at School of Medicine,University of California, Davis. He joined Department of Statistics, NorthCarolina State University, in 2005, as an Assistant Professor in Statistics,and was promoted to an Associate Professor with tenure in 2011. He was avisiting faculty at Department of Statistics, Stanford University and YahooResearch Labs from 2011 to 2013. He joined Division of Biostatistics,University of California, Berkeley, as an Associate Professor with tenure, in2014. His research interests include neuroimaging analysis, networks analysis,recommendation system, high dimensional regression, dimension reduction, machinelearning, and bioinformatics.
内容提要：
Classicalregression methods treat covariates as a vector and estimate a correspondingvector of regression coefficients. Modern applications in medical imaginggenerate covariates of more complex form such as multidimensional arrays(tensors). Traditional statistical and computational methods are provinginsufficient for analysis of these high-throughput data due to their ultrahighdimensionality as well as complex structure. In this talk, we propose a newfamily of tensor regression models that efficiently exploit the specialstructure of tensor covariates. Under this framework, ultrahigh dimensionalityis reduced to a manageable level, resulting in efficient estimation andprediction. A fast and highly scalable estimation algorithm is proposed formaximum likelihood estimation and its associated asymptotic properties arestudied. Effectiveness of the new methods is demonstrated on both synthetic andreal MRI imaging data.
