BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Probabilistic Graphical Models in Microsoft's Online Services: Tru
 eSkill\, AdPredictor\, and Matchbox - Thore Graepel\, MSR Cambridge
DTSTART:20111012T131500Z
DTEND:20111012T141500Z
UID:TALK32398@talks.cam.ac.uk
CONTACT:Stephen Clark
DESCRIPTION:Abstract:\nProbabilistic Graphical Models play a crucial role 
 in Microsoft's online services. In this talk\, I will describe three power
 ful applications of graphical model inference in practice. \n1.      TrueS
 kill is Xbox Live's Ranking and Matchmaking system and ensures that gamers
  online have balanced and exciting matches with equally skilled opponents.
  \n2.      AdPredictor is the system that estimates click-through rates (C
 TR) for ad selection and pricing within Microsoft's search engine Bing. \n
 3.      Matchbox is a large scale Bayesian recommender system that combine
 s aspects of collaborative filtering and content-based recommendation. It 
 is currently being used for tweet recommendation within\nprojectemporia.co
 m.\nAll three systems have in common that they are based on factor graph m
 odels and approximate Bayesian inference. They operate at a very large sca
 le involving millions of gamers\, billions of ad impressions\,\nand millio
 ns of tweets\, respectively. I will discuss the underlying graphical model
 s and inference algorithms as well as application-specific insights and fi
 ndings. Time permitting\, I will show the three\nsystems in action. This i
 s based on joint work with Ralf Herbrich\, David Stern\, Thomas Borchert\,
  Tom Minka\, and Joaquin Quiñonero Candela.\n
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR