BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Distributed\, Real-Time Bayesian Learning in Online Services - Ral
 f Herbrich (Facebook)
DTSTART:20121016T150000Z
DTEND:20121016T160000Z
UID:TALK40548@talks.cam.ac.uk
CONTACT:Eiko Yoneki
DESCRIPTION:The last ten years have seen a tremendous growth in Internet-b
 ased online services such as search\, advertising\, gaming and social netw
 orking.\nTraditionally\, offline analysis of large collections of user int
 eraction data has informed building predictive models for these services\;
  today\, adjusting and improving these models in real-time has become just
  as important.\n\nOne of the biggest challenges in this setting is scale: 
 not only does the sheer scale of data necessitate parallel processing but 
 it also necessitates distributed models\; with over 900 million active use
 rs at Facebook\, any user-specific sets of features in a linear or non-lin
 ear model yields models of a size bigger than can be stored in a single ph
 ysical computer.\n\nIn this talk\, I will give a hands-on introduction to 
 one of the most versatile tools for handling large collections of data wit
 h distributed probabilistic models: the sum-product algorithm for approxim
 ate message passing in factor graphs. I will discuss the application of th
 is algorithm for the specific case of generalized linear models and outlin
 e the challenges of both approximate and distributed message passing inclu
 ding an in-depth discussion of expectation propagation and the relation to
  Map-Reduce - a related technique for dealing with Big data and distribute
 d learning. The talk will be filled with experimental findings when runnin
 g such systems at the scale of Facebook.\n\nShort Bio: \nRalf is working a
 t Facebook on large-scale\, distributed learning and prediction as a web s
 ervices/infrastructure. Before joining Facebook\, he was heading the Bing 
 Personalization team which focused on prototyping and enabling personalize
 d experiences across Microsoft's Online Services Division. Prior this his 
 work on Bing\, Ralf was Director of Microsoft's Future Social Experiences 
 (FUSE) Labs UK working on new social experiences powered by computational 
 intelligence technologies on large online data collections. Ralf joined Mi
 crosoft Research in 2000 as a Postdoctoral researcher and Research Fellow 
 of the Darwin College Cambridge. During his time at Microsoft Research\, R
 alf was working in the areas of machine learning\, information retrieval\,
  game theory\, artificial intelligence\, optimization and social network a
 nalysis. Prior to joining Microsoft\, Ralf worked at the Technical Univers
 ity Berlin as a teaching assistant where he obtained both a Diploma degree
  in Computer Science and a Ph.D. degree in Statistics.\n\nRalf's research 
 interests include Bayesian inference and decision making\, computer games\
 , kernel methods\, statistical learning theory and distributed systems. Ra
 lf is one of the inventors of the Drivatars system used in the Forza Motor
 sport series as well as the TrueSkill ranking and matchmaking system in Xb
 ox 360 Live. He also co-invented the click-prediction technology used in B
 ing's online advertising system.\n
LOCATION:LT1\, Computer Laboratory\, William Gates Builiding
END:VEVENT
END:VCALENDAR