Professor of Biostatistics
Room 21-254A CHS
Department of Biostatistics
UCLA School of Public Health
Los Angeles, CA 90095-1772
- B.M. Clinical Medicine (2001) China Medical University
- M.S. Bioinformatics (2003) Iowa State University
- Ph.D. Statistics (2008) Stanford University
Statistical computing: Dr. Zhou has long term interests in numerical optimization problems, particularly those arising from statistical analysis of high-dimensional data. He developed highly scalable optimization algorithms for maximum likelihood estimation of some multivariate discrete distributions, calculation of importance sampling weights for large data sets, geometric and signomial programming, and a model-based movie rating method. He also proposed a new deterministic annealing method for global optimization, a quasi-Newton scheme for accelerating high-dimensional optimization algorithms, and a strategy for massive parallel computing using graphical processing units (GPUs). He studied new path following algorithms for regularization problems in statistics and machine learning, and successfully generalized them to least angle regression and convex programming. His recent development also includes scalable estimation algorithm for multivariate response generalized linear models and variance components models, fast matrix computation tools, and distance majorization for convex programming.
Neuroimaging: Modern technologies are producing a wealth of data with complex structures. For instance neuroimaging data often takes the form of multidimensional arrays, also known as tensors. Analysis of these high-throughput data calls for new regression and classification tools. Recently Dr. Zhou started methodology development for neuroimaging data analysis. With collaborators and students, he is actively extending their methods in various directions, including longitudinal image data and imaging genetics.
Statistical genetics/genomics: One of Dr. Zhou's research interests is to develop statistical and computational tools for analysis of large-scale genomic data. He developed penalization methods for association screening of genome-wide association (GWAS) and next generation sequencing (NGS) data, and a nonlinear dimension reduction approach for genotype aggregation and association mapping. Currently he is working on genome-wide QTL association mapping based on family designs, genotype imputation, transcriptomics data analysis based on RNA-seq technology, and statistical methods for analyzing microbiome data. Dr. Zhou is a developer of the comprehensive genetics analysis software Mendel, which is freely available at the UCLA Human Genetics Software Page. His current work includes implementing fast likelihood ratio test of variance components for genome wide QTL mapping and fast genotype imputation for pedigrees.
Markov chains and applications: Dr. Zhou's Ph.D research concerns convergence rates of Markov chains. With Dr. Mary Sehl et al at UCLA, he has developed a sequence of stochastic models for the dynamics of cancer stem cell population.