[Ml-stat-talks] Fwd: ORFE - Wilks Statistics Seminar: Pradeep Ravikumar, Univ. of TX, March 27th at 12:30 pm, Sherrerd Hall 101
bee at princeton.edu
Thu Mar 26 08:58:22 EDT 2015
Talk of great interest on Friday.
---------- Forwarded message ----------
From: Connie Brown <connieb at princeton.edu>
Date: Thu, Mar 26, 2015 at 8:55 AM
Subject: ORFE - Wilks Statistics Seminar: Pradeep Ravikumar, Univ. of TX,
March 27th at 12:30 pm, Sherrerd Hall 101
To: Barbara Engelhardt <bee at princeton.edu>
DATE: Friday, March 27, 2015
LOCATION: Sherrerd Hall, room 101
SPEAKER: Pradeep Ravikumar, University of Texas
TITLE: Elementary Estimators for High-dimensional Statistical Models
ABSTRACT: We consider the problem of learning high-dimensional statistical
models, where the number of variables could be potentially larger than the
number of observations. This class of problems has attracted considerable
attention over the last decade, with state of the art statistical
estimators based on solving regularized convex programs. Scaling these
typically non-smooth convex programs to the very large-scale problems of
the Big Data era comprises an ongoing and rich area of research.
In contrast to this two-stage approach of first devising statistically
efficient estimators, and then devising computationally efficient
optimization methods to solve these estimators, we attempt to address this
scaling issue at the source, by asking whether one can build simpler
closed-form estimators, that yet come with statistical guarantees that are
nonetheless comparable to regularized likelihood estimators. Surprisingly,
we answer this question in the affirmative. We analyze our estimators in
the high-dimensional setting, and moreover provide empirical corroboration
of their statistical and computational performance guarantees.
Joint work with Eunho Yang and Aurelie Lozano.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Ml-stat-talks