BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Automated Augmented Conjugate Inference for Gaussian Processes - T
 héo Galy-Fajou (TU Berlin)
DTSTART:20210420T130000Z
DTEND:20210420T140000Z
UID:TALK159358@talks.cam.ac.uk
CONTACT:Elre Oldewage
DESCRIPTION:Gaussian Processes are a tool of choice for modelling function
  with uncertainties. However\, inference is only tractable analytically  f
 or the classical case of regression with Gaussian noise since all other li
 kelihoods are not conjugate with the Gaussian prior.\n\nIn this talk I wil
 l show how one can transform a large class of likelihoods into conditional
  conjugate distributions by augmenting them with latent variables. These a
 ugmented models have the advantage that\, while the posterior inference is
  still not fully analytic\, the full conditionals are! Consequently\, one 
 can work easily (and efficiently!) with algorithms like Gibbs sampling or 
 Coordinate Ascent VI (CAVI) and outperform existing inference methods.\n\n
 Reference:\n\nGaly-Fajou\, Théo\, Florian Wenzel\, and Manfred Opper. "Au
 tomated Augmented Conjugate Inference for Non-conjugate Gaussian Process M
 odels." International Conference on Artificial Intelligence and Statistics
 . PMLR\, 2020. https://arxiv.org/abs/2002.11451
LOCATION:https://eng-cam.zoom.us/j/81961210430?pwd=cUZGbzU4NzJickd5THlsYzJ
 4cmlndz09
END:VEVENT
END:VCALENDAR