BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Département de mathématiques et applications - ECPv6.2.2//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Département de mathématiques et applications
X-ORIGINAL-URL:https://www.math.ens.psl.eu
X-WR-CALDESC:évènements pour Département de mathématiques et applications
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20100328T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20101031T010000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Paris:20100920T133000
DTEND;TZID=Europe/Paris:20100920T143000
DTSTAMP:20260413T182136
CREATED:20100920T113000Z
LAST-MODIFIED:20211104T085524Z
UID:7934-1284989400-1284993000@www.math.ens.psl.eu
SUMMARY:Unimodal bandits
DESCRIPTION:We consider multiarmed bandit problems where the expected reward isunimodal over a partially ordered set of arms. In particular\, thearms may belong to a continuous interval or correspond to verticesin a graph.  We present efficient algorithms to minimize the regretin these bandit problems and also to detect abrupt changes in thereward distributions.  The unimodality assumption has an importantadvantage: we can determine if a given arm is optimal by samplingthe possible directions around it. This property allows us toquickly find the optimal arm in a graph and detect changes. Notably\,our method incurs only a regret that depends logarithmically on thediameter of the graph.  \nLink to the paper
URL:https://www.math.ens.psl.eu/evenement/unimodal-bandits/
LOCATION:Ecole normale supérieure salle W
CATEGORIES:SMILE in Paris
END:VEVENT
END:VCALENDAR