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:Probabilistic Systems\, Information\, and Inferenc
 e Group Seminars
SUMMARY:Online temporally adaptive parameter estimation wi
 th applications to streaming data analysis. - Cris
 toforos Anagnostopoulos\, Statistics Laboratory\, 
 University of Cambridge
DTSTART;TZID=Europe/London:20101103T141500
DTEND;TZID=Europe/London:20101103T150000
UID:TALK27516AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/27516
DESCRIPTION:Online learning algorithms deployed in streaming d
 ata \ncontexts may be additionally required to pos
 sess temporally adaptive properties\, in order to 
 remain up-to-date against unforeseen changes\, smo
 oth or abrupt\, in the underlying data generation 
 mechanism. In cases where explicit dynamic modelli
 ng is either impossible or impractical\, temporall
 y adaptive behaviour may still be induced by contr
 olling the responsiveness of the estimator to nove
 l information.\nThis can be naturally accomplished
  in certain algorithms that feature user-specified
  learning rates\, such as the Robbins-Monro family
  of \nalgorithms. We discuss available methodology
  for automatic self-tuning learning rates in a Rob
 bins-Monro context. On the basis of both \ntheoret
 ical insights and real-data experiments\, we demon
 strate that this approach can efficiently handle t
 emporal variation of unknown characteristics\, whi
 le additionally serving as a monitoring tool.\n
LOCATION:LR10\, Engineering\, Department of
CONTACT:Rachel Fogg
END:VEVENT
END:VCALENDAR