BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Handling temporal variation of unknown characteristics in streamin
 g data analysis. - Chris Anagnostopoulos\, Cambridge University
DTSTART:20101130T140000Z
DTEND:20101130T150000Z
UID:TALK27981@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Data collection technology is undergoing a revolution that is 
 enabling streaming acquisition of real-time information in a wide variety 
 of settings. Faced with indefinitely long\, high frequency and possibly hi
 gh dimensional data sequences\, learning algorithms must rely on summary s
 tatistics and computationally efficient online inference without the need 
 to store and revisit the data history. Moreover\, learning must be tempora
 lly adaptive in order to remain up-to-date against unforeseen changes\, sm
 ooth or abrupt\, in the underlying data generation mechanism. In cases whe
 re explicit dynamic modelling is either impossible or impractical\, tempor
 ally adaptive behaviour may still be induced by controlling the responsive
 ness of the estimator to the novel information. We discuss ways in which t
 his can be accomplished in data-dependent manners for popular classes of o
 nline algorithms. We focus on the Robbins-Monro family of algorithms that 
 naturally feature a sequence of user-specified learning rates\, and discus
 s available methodology for automatic self-tuning in this context. On the 
 basis of both theoretical insights and real-data experiments\, we demonstr
 ate that such approaches can efficiently handle temporal variation of unkn
 own characteristics\, while additionally serving as a monitoring tool.
LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso
 n Avenue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR