BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:1423@i2m.univ-amu.fr
DTSTART;TZID=Europe/Paris:20161114T143000
DTEND;TZID=Europe/Paris:20161114T153000
DTSTAMP:20161030T133000Z
URL:https://www.i2m.univ-amu.fr/evenements/refined-lower-bounds-for-advers
 arial-bandits/
SUMMARY: (...): Refined lower bounds for adversarial bandits
DESCRIPTION:: We provide new lower bounds on the regret that must be suffer
 ed by adversarial bandit algorithms. The new results show that recent uppe
 r bounds that either (a) hold with high-probability or (b) depend on the t
 otal lossof the best arm or (c) depend on the quadratic variation of the l
 osses\, are close to tight. Besides this we prove two impossibility result
 s. First\, the existence of a single arm that is optimal in every round ca
 nnot improve the regret in the worst case. Second\, the regret cannot scal
 e with the effective range of the losses. In contrast\, both results are p
 ossible in the full-information setting. Page de Sébastien Gerchinovitz |
  preprint
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20161030T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
END:VTIMEZONE
END:VCALENDAR