BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Fitting Hierarchical Models in Large-Scale Recommender Systems - P
 atrick Perry (New York University)
DTSTART:20160804T130000Z
DTEND:20160804T140000Z
UID:TALK66944@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:Early in the development of recommender systems\, hierarchical
  models were recognized as a tool capable of combining content-based filte
 ring (recommending based on item-specific attributes) with collaborative f
 iltering (recommending based on preferences of similar users). However\, a
 s recently as the late 2000s\, many authors deemed the computational costs
  required to fit hierarchical models to be prohibitively high for commerci
 al-scale settings. This talk addresses the challenge of fitting a hierarch
 ical model at commercial scale by proposing a moment-based procedure for e
 stimating the parameters of a hierarchical model. This procedure has its r
 oots in a method originally introduced by Cochran in 1937. The method trad
 es statistical efficiency for computational efficiency. It gives consisten
 t parameter estimates\, competitive prediction error performance\, and sub
 stantial computational improvements. When applied to a large-scale recomme
 nder system application and compared to a standard maximum likelihood proc
 edure\, the method delivers competitive prediction performance while reduc
 ing computation time from hours to minutes.
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR