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:Isaac Newton Institute Seminar Series
SUMMARY:Consistency and CLTs for stochastic gradient Lange
vin dynamics based on subsampled data - Vollmer\,
S (University of Oxford)
DTSTART;TZID=Europe/London:20140424T155000
DTEND;TZID=Europe/London:20140424T162500
UID:TALK52165AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/52165
DESCRIPTION:Co-authors: Alexandre Thiery (National University
of Singapore)\, Yee-Whye Teh (University of Oxford
) \n\nApplying MCMC to large data sets is expensiv
e. Both calculating the acceptance probability and
creating informed proposals depending on the like
lihood require an iteration through the whole data
set. The recently proposed Stochastic Gradient La
ngevin Dynamics (SGLD) circumvents this problem by
generating proposals based on only a subset of th
e data and skipping the accept-reject step. In ord
er to heuristically justify the latter\, the step
size converges to zero in a non-summable way. \n\n
Under appropriate Lyapunov conditions\, we provide
a rigorous foundation for this algorithm by showi
ng consistency of the weighted sample average and
proving a CLT for it. Surprisingly\, the fraction
of the data subset selection does not have an infl
uence on the asymptotic variance.\n
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:Mustapha Amrani
END:VEVENT
END:VCALENDAR