Posted on Friday, March 26, 2021
This session is jointly sponsored by the Statistics department and the Research
Methods, Measurement, and Evaluation program, University of Connecticut
(UCONN), New England Statistical Society (NESS) and Statistical and Applied
Mathematical Institute (SAMSI) as part of online interdisciplinary seminar
series on statistical methodology for social and behavior research.
Speaker: DR. DAVID DUNSON, DUKE UNIVERSITY
Date and Time: FRIDAY, 3/26/2021, 12PM
Topic: Bayesian Pyramids: Identifying Interpretable Deep Structure
Underlying High-dimensional Data
Abstract:
High-dimensional categorical data are routinely collected in biomedical and
social sciences. It is of great importance to build interpretable models that
perform dimension reduction and uncover meaningful latent structures from such
discrete data. Identifiability is a fundamental requirement for valid modeling
and inference in such scenarios yet is challenging to address when there are
complex latent structures. We propose a class of interpretable discrete latent
structure models for discrete data and develop a general identifiability
theory. Our theory is applicable to various types of latent structures,
ranging from a single latent variable to deep layers of latent variables
organized in a sparse graph (termed a Bayesian pyramid). The proposed
identifiability conditions can ensure Bayesian posterior consistency under
suitable priors. As an illustration, we consider the two-latent layer model and
propose a Bayesian shrinkage estimation approach. Simulation results for this
model corroborate identifiability and estimability of the model parameters.
Applications of the methodology to DNA nucleotide sequence data uncover
discrete latent features that are both interpretable and highly predictive of
sequence types. The proposed framework provides a recipe for interpretable
unsupervised learning of discrete data and can be a useful alternative to
popular machine learning methods.
Bio:
Dr. David Dunson is Arts & Sciences Distinguished Professor of Statistical
Science and Mathematics at Duke University. His research focuses on developing
methodology for analysis and interpretation of complex and high-dimensional
data, with a particular emphasis on biomedical applications, Bayesian
statistics, and probability modeling approaches. Methods development and
theory is directly motivated by challenging applications in neuroscience,
genomics, environmental health, and ecology, among others. Dr. Dunson received
his BS in Mathematics from the Pennsylvania State University in 1994, and his
PhD in Biostatistics from Emory University in 1997. He then spent a decade at
the National Institute of Environmental Health Sciences before moving to
Duke. His work has had substantial impact, with ~55,000 citations on Google
Scholar and an H-index of 80.
For inquiry, please contact Dr. Xiaojing Wang at xiaojing.wang@uconn.edu
