BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Gaussian processes with neural network inductive biases for fast d
 omain adaptation - Andreas Damianou (Amazon)
DTSTART:20200228T173000Z
DTEND:20200228T190000Z
UID:TALK140116@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Recent advances in learning algorithms for deep neural network
 s allows us to train such models efficiently to obtain rich feature repres
 entations.  However\, when it comes to transfer learning\, local optima in
  the high-dimensional parameter space still pose a severe problem. Motivat
 ed by this issue\, we propose a framework for performing probabilistic tra
 nsfer learning in the function space while\, at the same time\, leveraging
  the rich representations offered by deep neural networks. Our approach co
 nsists of linearizing neural networks to produce a Gaussian process model 
 with covariance function given by the network's Jacobian matrix. The resul
 t is a closed-form probabilistic model which allows fast domain adaptation
  with accompanying uncertainty estimation.  
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR