Topic one: Sparse Estimation via Approximated Information Criteria
Speaker： Dr. Xiaogang Su
Abstract:
We propose a new
method for sparse estimation by directly optimizing an approximated information
criterion. The main idea involves approximation of the l0 norm with
a continuous or smooth unit dent function (as exemplified by the hyperbolic
tangent function). The proposed method bridges the best subset selection and
regularization by borrowing strength from both; it mimics best subset selection
using a penalized likelihood approach yet with no need of a tuning parameter.
We fur-
ther reformulate
the problem with a reparameterization step so that it reduces to one
unconstrained nonconvex yet smooth programming problem, which can be solved efficiently
as in computing MLE. The reparameterization tactic yields an additional
advantage in circumventing post-selection inference. The asymptotic properties
of the proposed method are explored for both fixed and diverging dimensions.
Both simulated experiments and empirical examples are provided for assessment
and illustration.
Time：9:00--10:00am, July 3, 2015
Room：X2511
Profile：Dr. Xiaogang Su is an associate Professor at Department of
Mathematical Sciences, University of Texas at El Paso (UTEP). He got his Ph.D.
in Statistics from University of California at Davis in 2001. He has authored
more than 60 scientific research articles in indexed journals, refereed
conferences and books. He has been an associate editor for Journal of
Computational and Graphical Statistics (JCGS), (2010–Present) and Editorial
Board of Nursing Research (2012-Present).
Topic two: Statistical aggregation in big
data
Speaker： Dr. Nan Lin
Abstract:
Big data problems present great challenges to statistical
analyses, especially from the computational side. We consider a wide range of
statistical inference problems in big data problems. The statistical aggregation
strategy is a divide-and-conquer approach that aims to achieve asymptotic
equivalence. In addition to solve memory and storage difficulties appeared in
big data, it may also provide a computational efficient strategy in a non-big
data context. Through both theoretical proof and simulations, we show
that our method significantly reduces the computational time and meanwhile
maintains the asymptotic efficiency.
Time：10:15--11:15, July 3, 2015
Room：X2511
Profile：Dr.
Nan Lin is an associate
Professor at Department of Mathematics Washington University in St. Louis (USA).
He got his Ph.D. from Department of Statistics, University of Illinois at
Urbana-Champaign in 2003. His research interests lies in Statistical computing
in massive data, bioinformatics, Bayesian regularization, longitudinal and
functional data analysis, and statistical applications in anesthesiology and
cognition. He has been as associate editor for J. Computational Statistics and
Data Analysis (2011-present).
