BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:An Instability in Variational Methods for Learning Topic Models - 
 Andrea Montanari (Stanford University)
DTSTART:20180116T090000Z
DTEND:20180116T094500Z
UID:TALK97627@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:Topic models are extremely useful to extract latent degrees of
  freedom form large unlabeled datasets. Variational Bayes algorithms are t
 he approach most commonly used by practitioners to learn topic models. The
 ir appeal lies in the promise of reducing the problem of variational infer
 ence to an optimization problem. I will show that\, even within an idealiz
 ed Bayesian scenario\, variational methods display an instability that can
  lead to misleading results. [Based on joint work with Behroz Ghorbani and
  Hamid Javadi]
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR