BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Matrix Factorization and Relational Learning - Ajit Paul Singh (CM
 U)
DTSTART:20080909T130000Z
DTEND:20080909T140000Z
UID:TALK13396@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Matrix factorization is one of the workhorse methods in data m
 ining\, machine learning\, and information retrieval. We present a unified
  view of matrix factorization models\, which includes weighted singular va
 lue decompositions\, non-negative matrix factorization\, probabilistic lat
 ent semantic indexing\, max-margin matrix factorization\, matrix co-cluste
 ring\, and generalizations of these models to exponential family distribut
 ions. This unified view leads to a class of optimization algorithms\, base
 d on alternating projections and stochastic approximations\, which are wel
 l-suited to models of large\, sparse matrices.\n\nExtending upon our unifi
 ed view of matrix factorization\, many types of relational data can be pre
 sented as a set of related matrices\, where shared dimensions correspond t
 o shared factors in a low-rank representation. We extend Bregman matrix fa
 ctorization to a set of related matrices\, illustrating the use of relatio
 nal learning on a collaborative filtering problem.\n\nThis talk is based p
 rimarily on two publications: _Relational Learning via Collective Matrix F
 actorization_ (Singh & Gordon\, KDD-2008)\, and _A Unified View of Matrix 
 Factorization Models_ (Singh & Gordon\, ECML/PKDD-2008).\n\n
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR