统计学系系列讲座之241-242期

统计学系系列讲座之241期

时间：2016年11月22日（星期二）下午13:30—14:30

地点：史带楼504室

主持人：张新生 教授 复旦大学管理学院统计学系

主题：Hidden Markov Models on Variable Blocks with a Modal Clustering Algorithm and Applications

主讲人：Prof.Jia Li    Professor of Statistics, Penn State University

摘要：We develop a new hierarchical mixture model, namely Hidden Markov Model on Variable Blocks (HMM-VB), and a new mode search algorithm called Modal Baum-Welch (MBW) for efficient clustering of large-scale data that contain multiple rare clusters. Our work is motivated by high-throughput single-cell cytometry data with applications to vaccine development and immunological research. The curse of dimensionality and the highly unbalanced cluster structure of such data pose great difficulty for the usual framework of mixture models. Exploiting the widely accepted chain-like dependence among groups of variables in the single-cell cytometry data, we propose to treat the hierarchy of variable groups as a figurative time line and employ a HMM-type model, namely HMM-VB. Because component-wise clustering of data is unviable for HMM-VB, we propose to use mode-based clustering, aka modal clustering, and overcome the exponential computational complexity by MBW. We conduct a series of experiments on simulated data to demonstrate the power of HMM-VB and MBW, and make comparisons with existing methods. We also apply our method to identify rare cell subsets in cytometry data and examine its strengths and limitations.

统计学系系列讲座之242期

时间：2016年11月22日（星期二）下午14:30-15:30

地点：史带楼504室

主持人：张新生 教授 复旦大学管理学院统计学系

主题：Posterior contraction rates for deconvolution of Dirichlet-Laplace mixtures

主讲人：高凤楠 博士   复旦大学大数据学院、上海数学中心

报告人简介：Fengnan Gao is the new assistant professor jointly appointed

by School of Data Science and Shanghai Center for Mathematical Science. 

His completed his PhD in Leiden University under the guidance of Aad van

der Vaart. His research focuses on nonparametric Bayesian statistics,

statistical inference in network science, probabilistic methods in

complex networks and modeling and analysis of social networks.

摘要：We study nonparametric Bayesian inference with location mixtures of the Laplace density and a Dirichlet process prior on the mixing distribution. We derive a contraction rate of the corresponding posterior distribution, both for the mixing distribution relative to the Wasserstein metric and for the mixed density relative to the Hellinger and Lq metrics.

统计学系

2016-11-17