当前位置:首页 > 学术报告 > 2011年学术报告

报告题目: MULTIPLE TESTING VIA FDR_L FOR LARGE-SCALE IMAGING DATA
报 告 人: 冯进教授,美国堪萨斯大学
时间地点: 2011年7月26-28,30日下午3:00 S712
摘       要:

In this series of lectures, I will first introduce two classes of first order Hamilton-Jacobi-Bellman (HJB) equations in the space of probability measures, and then develop a well posed-ness theory for both the Cauchy and resolvent type problems.
In the first lecture, I will derive two Hamiltonians using probabilistic large deviation theory applied to mean-field models from statistical mechanics, provide their connections with optimal controlled PDEs in space of measures, and explain some a priori estimates at microscopic level using probability theory.
In the second lecture, I will introduce necessary background about optimal mass transportation theory and use the language of such theory to prove uniqueness (comparison principle) result for HJB equations involving one of the Hamiltonians (the nicer one).
In the third lecture, I will improve the comparison principle result in the last lecture by regularization method and obtain continuity result of the solution under a reasonably weak topology (the Wasserstein-2 metric).
In the fourth lecture, I will introduce new ingredients to show the uniqueness (comparison principle) of HJB equations for the second Hamiltonian which has a strong singular term. I will also explain the relation of such equation to the Onsager-Joyce-Montgomery theory of 2-D turbulent vortex flows.

报告人简历:  
 
报告题目: MULTIPLE TESTING VIA FDR_L FOR LARGE-SCALE IMAGING DATA
报 告 人: Prof. Chunming Zhang,University of Wisconsin
时间地点: 2011年7月25日上午10:00 S1013
摘       要:

The multiple testing procedure plays an important role in detecting the presence of spatial signals for large-scale imaging data. Typically, the spatial signals are sparse but clustered. This paper provides empirical evidence that for a range of commonly used control levels, the conventional FDR procedure can lack the ability to detect statistical significance, even if the p-values under the true null hypotheses are independent and uniformly distributed; more generally, ignoring the neighboring information of spatially structured data will tend to diminish the detection effectiveness of the FDR procedure. This paper first introduces a scalar quantity to characterize the extent to which the “lack of identification phenomenon” (LIP) of the FDR procedure occurs.Second, we propose a new multiple comparison procedure, called FDR_L, to accommodate the spatial information of neighboring p-values, via a local aggregation of p-values. Theoretical properties of the FDR_L procedure are investigated under weak dependence of p-values. It is shown that the FDR_L procedure alleviates the LIP of the FDR procedure, thus substantially facilitating the selection of more stringent control levels. Simulation evaluations indicate that the FDR_L procedure improves the detection sensitivity of the FDR procedure with little loss in detection specificity. The computational simplicity and detection effectiveness of the FDR_L procedure are illustrated through a real dataset.

报告人简历:  
 
报告题目: Semiparametric estimation under length-biased sampling
报 告 人: Dr. CHIUNG-YU HUANG,Biostatistics Research Branch,National Institute of Allergy and Infectious Diseases (NIAID),National Institutes of Health (NIH)
时间地点: 2011年7月13日下午4:00 S703
摘       要:

The prevalent cohort sampling design is frequently adopted to study the natural history of a disease. Important examples include epidemiological studies of disease etiology (Simon, 1980), cancer screening trials (Zelen and Feinleib, 1969), and HIV prevalent cohort studies (Lagakos et al, 1988). In addition to right censoring due to loss to follow-up, survival data collected via prevalent sampling are subject to left truncation, as those who fail before the recruitment time are not observable.In this talk, I will present a pseudo-profile likelihood method for survival data arising from length-biased sampling, where the survival times are right censored and left truncated by uniformly distributed random truncation times. The proposed method does not involve estimation of the censoring distribution; hence it is consistent even when the censoring time is correlated with covariates. Simulation studies show that the proposed estimator yields significant improvement in efficiency over the popular maximum partial likelihood estimator. To illustrate the proposed method, statistical analyses using data from the Canadian Study of Health and Aging (CSHA)and the CDC AIDS Blood Transfusion Study will be reported. (This is joint work with Drs. Jing Qin and Dean Follmann.)

报告人简历:  
 
报告题目: A multiscale variational optimization problem with dynamical partial differential equation constraints for applications of atmospheric data assimilation
报 告 人: 谢元富 研究员,National Oceanic and Atmospheric Administration (NOAA)
时间地点: 2011年4月20日下午3:30 S703
摘       要:

Variational atmospheric data assimilation is a variational minimization problem subject to a set of nonlinear partial differential equations and is very important for improving numerical weather prediction at various time scales from tornados to hurricanes. It is very large and complex optimization problem. In this presentation, we review a few important features of a Space and Time Multiscale Analysis System (STMAS) developed at NOAA Earth System Research Laboratory for improving variational data assimilation and forecasts. We also review several outstanding challenges in general data assimilation in variational or Kalman filter. Some applications and numerical results will be shown

报告人简历: 谢元富研究员自1991年以来在美国国家海洋气象管理局的地球系统研究所从事数据同化和预报模式的研究。现任职称是研究员,目前担任若干项目的负责人,如,STMAS开发及应用项目,美国航空管理局暴风边界识别项目,NOAA观察系统模拟试验系统项目OSSE以及NOAA太平洋观察研究及预报试验项目。目前研究重点是其他在2000年提出的"时空多尺度同化系统(STMAS)"的开发及应用。STMAS是一个以微分方程为约束条件的大规模变分优化系统。控制变量个数在十的七到八次方以上。
 
报告题目: Differential Harnack inequality on Kaehler manifolds
报 告 人: Prof. Lei Ni, University of California at San Diego
时间地点: 2011年3月8日下午3:00 S712
摘       要:

 

 
当前位置:首页 > 学术报告 > 2011年学术报告