BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//128.220.117.42//NONSGML kigkonsult.se iCalcreator 2.20//
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-FROM-URL:https://www.clsp.jhu.edu
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:STANDARD
DTSTART:20171105T020000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20180311T020000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:5061-1426824000-1426910399@www.clsp.jhu.edu
DTSTAMP:20180319T112035Z
CATEGORIES;LANGUAGE=en-US:Seminars
CONTACT:
DESCRIPTION:Abstract\nIn many practical scenarios\, complex high-dimensiona
l data contains low-dimensional structures that could be informative of th
e analytic problems at hand. I will present a method that detects such str
uctures if they exist\, and uses them to construct compact interpretable m
odels for different machine learning tasks that can benefit practical appl
ications.\nTo start with\, I will formalize Informative Projection Recover
y\, the problem of extracting a small set of low-dimensional projections o
f data that jointly support an accurate model for a given learning task. O
ur solution to this problem is a regression-based algorithm that identifie
s informative projections by optimizing over a matrix of point-wise loss e
stimators. It generalizes to multiple types of machine learning problems\,
offering solutions to classification\, clustering\, regression\, and acti
ve learning tasks. Experiments show that our method can discover and lever
age low-dimensional structures in data\, yielding accurate and compact mod
els. Our method is particularly useful in applications in which expert ass
essment of the results is of the essence\, such as classification tasks in
the healthcare domain.\nBiography\nMadalina Fiterau is a PhD student in M
achine Learning at Carnegie Mellon University and a member of the Auton La
b. She is advised by Prof. Artur Dubrawski. Her research interests include
query-specific models for decision support systems\, learning with struct
ured sparsity\, dimensionality reduction in an active learning setting and
anomalous pattern detection. She received her BE in Computer Engineering
from the Politehnica University of Timisoara.
DTSTART;VALUE=DATE:20150320
DTEND;VALUE=DATE:20150321
SEQUENCE:0
SUMMARY:Ensembles for the Discovery of Compact Structures in Data – Madalin
a Fiterau (Carnegie Mellon University)
URL:https://www.clsp.jhu.edu/events/ensembles-for-the-discovery-of-compact-
structures-in-data-madalina-fiterau-carnegie-mellon-university/
X-ALT-DESC;FMTTYPE=text/html:\\n\\n\\n\\n\\nAbstract

\nIn many practical scenarios\, complex high-dimensional data contains lo
w-dimensional structures that could be informative of the analytic problem
s at hand. I will present a method that detects such structures if they ex
ist\, and uses them to construct compact interpretable models for differen
t machine learning tasks that can benefit practical applications.

\nT
o start with\, I will formalize Informative Projection Recovery\, the prob
lem of extracting a small set of low-dimensional projections of data that
jointly support an accurate model for a given learning task. Our solution
to this problem is a regression-based algorithm that identifies informativ
e projections by optimizing over a matrix of point-wise loss estimators. I
t generalizes to multiple types of machine learning problems\, offering so
lutions to classification\, clustering\, regression\, and active learning
tasks. Experiments show that our method can discover and leverage low-dime
nsional structures in data\, yielding accurate and compact models. Our met
hod is particularly useful in applications in which expert assessment of t
he results is of the essence\, such as classification tasks in the healthc
are domain.

\nBiography

\nMadalina Fiterau is a PhD student in M
achine Learning at Carnegie Mellon University and a member of the Auton La
b. She is advised by Prof. Artur Dubrawski. Her research interests include
query-specific models for decision support systems\, learning with struct
ured sparsity\, dimensionality reduction in an active learning setting and
anomalous pattern detection. She received her BE in Computer Engineering
from the Politehnica University of Timisoara.

\n\n