報(bào)告1

報(bào)告人：Associate Prof. Yiyuan She

講座地點(diǎn)：2015年6月25日上午8:30-9:30

講座地點(diǎn)：犀浦校區(qū)X2511

 主講人簡介：

Yiyuan She is an associate professor at Department of Statistics,Florida State University (USA).He received his Ph.D. in statistics fromStanfordUniversity (USA) in 2008. His research interests lies in the fields of High Dimensional Statistics, Statistical Machine Learning, Statistical Computing, Optimization, Multivariate Statistics, Robust Statistics, Network Science. He has been a member of Institute of Mathematical Statistics (IMS) , and a member of American Statistical Association (ASA).

 講座內(nèi)容簡介：

Talk Title: Oracle Inequalities for $\Theta$-estimators (關(guān)于$\Theta$-估計(jì)量的甲骨文不等式)

Abstract: Due to the explosion of large-scale datasets in statistical applications, people often favor first-order optimization methods to obtain an estimator in complex learning tasks. This talk performs statistical analysis of a class of thresholding (denoted by $\Theta$) based estimators defined in this way. They can be associated with a wide family of sparsity-inducing penalties but do not guarantee local optimality. Oracle inequalities are shown to provide nearly minimax rate optimality of $\Theta$-estimators under various conditions. We also prove that the sequence of iterates can decay to the desired statistical accuracy geometrically fast. Our results reveal different benefits brought by convex and nonconvex types of shrinkage.

報(bào)告2

題目: Analysis and Experimental Designs for Determing Benchmark Dosages with Multiple Endpoints and Multiple Stressors

報(bào)告人：Dr. Edward Boone

報(bào)告內(nèi)容:

Toxicologists are often faced with determining the amount of a substance or stressor that a person may be exposed to before adverse effects happen on some health outcomes (endpoints).  Traditional methodology uses a univariate approach with a single chemical at a time.  In this work we consider not only multiple stressors but also multiple endpoints to determine a benchmark dose tolerable region for the combination of stressors that simultaneously considers all endpoints.  In addition develop and criterion for determining which endpoints are more sensitive than others.  Due to the complexity of the question we also explore various experimental design criteria for follow up experimentation.  

時(shí)間：2015年6月25日上午9:45--10:45

地點(diǎn)：犀浦校區(qū)2511

主講人簡介：Dr.Edward Boone is an associate professor at Department of Statistical Sciences and Operations Research, Virginia Commonwealth University (USA).  His research interests mainly lies inBayesian statistics, hierarchical models, missing data, model selection, uncertainty quantification, ecological and health applications. Dr. Boone serves as reviewers of Journal of Statistical Computation and Simulation，Transactions on Modelling and Computer Simulation，Computational Statistics and Data Analysis etc.

報(bào)告3

題目: Dimension reduction based on the Hellinger integral

報(bào)告人：Dr. Qin Wang

報(bào)告內(nèi)容:

Sufficient dimension reduction (SDR) is a useful tool to study the dependence between a response and a multidimensional predictor. A new formulation is proposed based on the Hellinger integral of order two, introduced as a natural measure of the regression information contained in a predictor subspace. The response may be either continuous or discrete. The link between local and global central subspaces is established. Relative to existing methods, its overall performance is broadly comparable. Computationally, it is very efficient, allowing larger problems to be tackled. (This is a joint work with Prof. Xiangrong Yin at The University of Kentucky and Prof. Frank Critchley at The Open University of UK.)

時(shí)間：2015年6月25日上午11:00--12:00

地點(diǎn)：犀浦校區(qū)2511

主講人簡介：Dr.Qin Wang is an associate professor at Department of Statistical Sciences and Operations Research, Virginia Commonwealth University (USA). He received his Ph.D in statistics from the University of Georgia in 2009. His research interests lies in Dimension reduction and variable selection, Multivariate statistical methods, Nonparametric/Semiparametric statistic methods, Statistical applications. He has been an associate editor for J. Statistics and Probability Letters since 2015.

主辦：研究生院

承辦：kaiyun開云官方網(wǎng)站