BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sample completion\, Netflix Prize Competition and k-dependence - L
 eonardo Coregliano (University of Chicago)
DTSTART:20260114T144500Z
DTEND:20260114T153000Z
UID:TALK241633@talks.cam.ac.uk
CONTACT:Julia Wolf
DESCRIPTION:In the Netflix Prize Competition (2006-2009)\, we are given a 
 finite set U of users\, a finite set M of movies and a partial\nfunction f
 : U x M ⇀ R\, where f(u\,m) indicates how user u rates movie m and we ar
 e tasked with\ncompleting f to a total function to predict how all users U
  rate all movies in M. Although some algorithms did fairly well\nin the co
 mpetition\, giving a satisfactory theoretical explanation for their succes
 s has been difficult. \n\nIn this second talk of\nthe series on high-arity
  learning frameworks\, I will discuss how the Netflix Prize Competition ca
 n be seen as an instance of a\nhigh-arity learning framework called "sampl
 e completion learning". I will also discuss how sample completion learning
  is\ncompletely characterized by the model-theoretic notion of k-dependenc
 e introduced by Shelah (which can be seen as a\nhigh-dimensional version o
 f the Vapnik--Chervonenkis dimension). In turn\, this gives a full theoret
 ical characterization of when\nthe Netflix problem is "solvable".\n\nNo ba
 ckground in learning theory or model theory is required for this talk.\n\n
 This talk is based on joint work with Maryanthe Malliaris.
LOCATION:MR13\, CMS
END:VEVENT
END:VCALENDAR