BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//6.4.8//EN
TZID:America/New_York
X-WR-TIMEZONE:America/New_York
BEGIN:VEVENT
UID:345@fds.yale.edu
DTSTART;TZID=America/New_York:20240327T113000
DTEND;TZID=America/New_York:20240327T130000
DTSTAMP:20240312T002814Z
URL:https://fds.yale.edu/events/fds-colloquium-aaditya-ramdas-cmu/
SUMMARY:FDS Colloquium: Aaditya Ramdas (CMU)\, "The numeraire e-variable an
d reverse information projection"
DESCRIPTION:\n\n\n\n\nSpeaker: Aaditya RamdasAssistant Professor\, Departme
nt of Statistics &\; Data Science (75%)\,Machine Learning Department (2
5%)\,Carnegie Mellon UniversityWednesday\, March 27\, 2024Lunch: 11:30 am
(Kitchen)Talk: 12:00 pm (Seminar Room #1327)at the Yale institute for Fou
ndations of Data Science\, Kline Tower\, 13th Floor\n\n\nTitle: The numera
ire e-variable and reverse information projection\n\n\nAbstract: In an exc
ellent 1999 Yale PhD thesis\, Jonathan Li proposed and defined a critical
concept that he called the reverse information projection (RIPr)\, which i
s akin to a KL projection of a distribution onto a set of probability meas
ures. This concept has gained prominence recently in game-theoretic statis
tics and sequential testing by betting\, because it characterizes the log-
optimal bet/e-variable of a point alternative hypothesis against a composi
te null hypothesis. However\, it required assumptions of convexity of the
set of distributions and a common reference measure to define densities. I
n this talk\, we will show how to fully and completely generalize the theo
ry underlying the RIPr\, showing that it is always well defined\, without
any assumptions on the distributions involved. Further\, a strong duality
result identifies it as the dual to an optimal bet/e-variable called the n
umeraire\, which is unique and also always exists without assumptions. Thi
s fully generalizes Kelly betting to composite nulls\, and also results by
Grunwald and coauthors on safe testing. The talk will not assume any prio
r knowledge on these topics. \n\nThis is joint work with Martin Larsson an
d Johannes Ruf (https://arxiv.org/abs/2402.18810).\n\nBio: Aaditya Ramdas
(PhD\, 2015) is an assistant professor at Carnegie Mellon University\, in
the Departments of Statistics and Machine Learning. His research interests
include game-theoretic statistics and sequential anytime-valid inference\
, multiple testing and post-selection inference\, and uncertainty quantifi
cation for machine learning (conformal prediction\, calibration). His appl
ied areas of interest include neuroscience\, genetics and auditing (real-e
state\, finance\, elections). Aaditya received the IMS Peter Gavin Hall Ea
rly Career Prize\, the COPSS Emerging Leader Award\, the Bernoulli New Res
earcher Award\, the NSF CAREER Award\, the Sloan fellowship in Mathematics
\, and faculty research awards from Adobe and Google. He also spends 20% o
f his time at Amazon working on causality and sequential experimentation.\
nWebsite: https://www.stat.cmu.edu/~aramdas/\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\n\n\n\n\n\n\n\n\n\n\n\n\n\n
CATEGORIES:Colloquium,Seminar Series
LOCATION:Yale Institute for Foundations of Data Science\, Kline Tower 13th
Floor\, Room 1327\, New Haven\, CT\, 06511\, United States
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Kline Tower 13th Floor\, Ro
om 1327\, New Haven\, CT\, 06511\, United States;X-APPLE-RADIUS=100;X-TITL
E=Yale Institute for Foundations of Data Science:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
DTSTART:20240310T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR