BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Statistics
SUMMARY:The information complexity of sequential resource
allocation - Emilie Kaufmann (INRIA LIlle)
DTSTART;TZID=Europe/London:20160422T160000
DTEND;TZID=Europe/London:20160422T170000
UID:TALK65792AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/65792
DESCRIPTION:This talk will be about sequential resource alloca
tion\, under the so-called stochastic multi-armed
bandit model. In this model\, an agent interacts w
ith a set of (unknown) probability distributions\,
called 'arms' (in reference to 'one-armed bandits
'\, another name for slot machines in a casino). W
hen the agent draws an arm\, he observes a sample
from the associated distribution. This sample can
be seen as a reward\, and the agent then aims at m
aximizing the sum of his rewards during the intera
ction. This 'regret minimization' objective makes
sense in many practical applications\, starting wi
th medical trials\, that motivated the introductio
n of bandit problems in the 1930's. Another possi
ble objective for the agent\, called best-arm iden
tification\, is to discover as fast as possible th
e best arm(s)\, that is the arms whose distributio
ns have highest mean\, but without suffering a los
s when drawing 'bad' arms. \n \nFor each of these
objectives\, our goal will be to define a distribu
tion-dependent notion of optimality\, thanks to lo
wer bounds on the performance of good strategies\,
and to propose algorithms that can be qualified a
s optimal according to these lower bounds. For som
e classes of parametric bandit models\, this permi
ts to characterize the complexity of regret minimi
zation and best-arm identification in terms of (di
fferent) information-theoretic quantities.
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberfo
rce Road\, Cambridge.
CONTACT:Quentin Berthet
END:VEVENT
END:VCALENDAR