BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium : Probabilistic representation learning and scalable
  Bayesian inference
DTSTART:20171012T161500
DTEND:20171012T173000
DTSTAMP:20260407T162255Z
UID:db3bfdce0aa59704867b0c4a4deed50145b6fd81c1211d82778435a8
CATEGORIES:Conferences - Seminars
DESCRIPTION:By : Stephan Mandt - Disney Research Pittsburgh\nVideo of his 
 talk\n\nAbstract : Probabilistic modeling is a powerful paradigm which has
  seen dramatic innovations in recent years. These innovations in approxima
 te inference\, mainly due to automatic differentiation and stochastic opti
 mization\, have made probabilistic modeling scalable and broadly applicabl
 e to many complex model classes. I start my talk by reviewing the dynamic 
 skip-gram model (ICML 2017) as an example of this class. The model results
  from combining a probabilistic interpretation of word embeddings with lat
 ent diffusion priors\, and allows us to study the dynamics of word embeddi
 ngs for text data that are associated with different time stamps. Our Baye
 sian approach allows us to share information across the time domain\, and 
 is robust even when the data at individual points in time is small. As a r
 esult\, we can automatically detect words that change their meanings even 
 in moderately-sized corpora. Yet\, the model is Bayesian non-conjugate\, a
 nd therefore we have to draw on modern variational inference methods to tr
 ain it efficiently on large data. The second part of my talk is therefore 
 devoted to advances in variational inference. Here\, I will review our ver
 y recent perturbative black box variational inference algorithm (NIPS 2017
 )\, that uses variational perturbation theory of statistical physics to co
 nstruct corrections to the standard variational lower bound. Last\, I will
  demonstrate that simple stochastic gradient descent with a constant step 
 size is a form of approximate Bayesian inference (JMLR and ICML 2016). \n
 \nBio : Stephan Mandt is a Research Scientist and head of the statistical 
 machine learning group at Disney Research Pittsburgh\, co-located with CMU
 . Previously\, he was a postdoctoral researcher with David Blei at Columbi
 a University (2014-2016)\, and a PCCM postdoctoral fellow at Princeton Uni
 versity (2012-2014). Stephan Mandt holds a Ph.D. in theoretical physics fr
 om the University of Cologne\, supported by the German National Merit Foun
 dation. His interests include large-scale probabilistic modeling\, represe
 ntation learning\, variational inference\, and media analytics.\n\nMore in
 formation    
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR