[Ml-stat-talks] Fwd: [ORFE seminars] ORFE Colloquium: David Dunson, Today at 4:30pm, Sherrerd Hall 101
bee at princeton.edu
Tue Oct 13 09:26:39 EDT 2015
Talk of interest today.
---------- Forwarded message ----------
From: Carol Smith <carols at princeton.edu>
Date: Tue, Oct 13, 2015 at 9:19 AM
Subject: [ORFE seminars] ORFE Colloquium: David Dunson, Today at 4:30pm,
Sherrerd Hall 101
To: ORFE-TALKS at princeton.edu
=== ORFE Colloquium Announcement ===
DATE: Today, October 13, 2015
LOCATION: Sherrerd Hall, room 101
SPEAKER: David Dunson, Duke University
TITLE: Probabilistic Modeling of High-dimensional Tables and Networks
ABSTRACT: High-dimensional discrete data are collected in many
application areas, but have seen limited consideration in the literature.
We focus in particular on two related problems: (i) high-dimensional
categorical data that can be organized as a many way contingency table; and
(ii) graph/network-valued data consisting of a binary adjacency matrix
measured for each subject. The first problem arises in numerous
applications ranging from survey research in social sciences and
epidemiology to genomics and marketing. The second arises in analysis of
brain connectomes and team dynamics among others. We propose a general
approach for these problems based on probabilistic tensor factorizations,
which are low rank and sparse. Taking a Bayesian nonparametric approach to
inference, we show large support, posterior consistency and convergence
rates in the high-dimensional setting. Efficient Markov chain Monte Carlo
algorithms are developed for practical implementation, and the methods are
illustrated through applications in epidemiology and neuroscience.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Ml-stat-talks