时间：2019年5月24日14：00―16：15

地点：玉泉校区数学中心五楼报告厅

a)Title：Shape-constrained Inference: Monotonicity of Densities

Time: 14:00―14:45 

Speaker: Prof. YAM, Phillip（CUHK）

Abstract: An a priori understanding of the (even just topological) structure of the data from a specific 

discipline can often strengthen much of the effectiveness of the inference, both for estimation and hypothesis  
testing, behind the mechanism of interest. Although the celebrated work of Grenander (1956) on the estimation  
of decreasing densities has been around for long, it is until recently that Shape-constrained inference has 
 become one of the most popular research areas in Statistics. One major reason behind is due to the 
challenges and mathematical difficulties from the theory; indeed, just limit to the scope of monotonicity of 
densities, there still remain plenty of lasting problems. In this talk, I shall introduce a couple of them, and then 
discuss on what the final resolution we provided. 

b)Title&abstract: to be confirmed

Time: 14:45―15:30 

Speaker: Prof. KONG Dexing

c)Title: Cluster analysis and visualization of sparse categorical data

Time: 15:30―16:15

Speaker: Prof. ZHANG Peng

Abstract: Sparsity in features presents a big technical challenge to existing cluster analysis and visualization of categorical data. Hierarchical Bayesian Bernoulli mixture model (HBBMM) incorporates constrained empirical Bayes priors for model parameters, so the resulting Expectation Maximization (EM) algorithm of estimator searching is confined in a proper region. The EM algorithm enables to obtain the maximum a posterior (MAP) estimation, in which cluster labels are simultaneously assigned. Another mixture model-based approach to clustering categorical data, as well as visualization method using latent variables, is also discussed. Several real-world sparse categorical datasets are analyzed with the proposed methods. 

联系人：王凤仪（wangfy@zju.edu.cn）