BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sequential Monte Carlo and deep regression - Thomas Schön\, Uppsa
 la University
DTSTART:20200130T140000Z
DTEND:20200130T150000Z
UID:TALK131410@talks.cam.ac.uk
CONTACT:Alberto Padoan
DESCRIPTION:This talk has two (for now) loosely connected parts: In the fi
 rst part we aim to provide intuition for the key mechanisms underlying the
  sequential Monte Carlo (SMC) method (including the popular particle filte
 rs and smoothers). SMC provide approximate solutions to integration proble
 ms where there is a sequential structure present. The classical example of
  such a structure is offered by nonlinear dynamical systems\, but we stres
 s that SMC is significantly more general than most of us first thought. We
  will hint at a few ways in which SMC fits into the machine learning toolb
 ox and mention a few interesting avenues for research. In the second part 
 we develop a new approach to deep regression. While deep learning-based cl
 assification is generally addressed using standardized approaches\, a wide
  variety of techniques are employed when it comes to regression. We have d
 eveloped a new and general deep regression method with a clear probabilist
 ic interpretation. We obtain good performance on several computer vision r
 egression tasks (including a new state-of-the-art result on visual trackin
 g). The loose connection lies in the use of the Monte Carlo idea in both t
 opics. We do believe that the connection between the two seemingly dispara
 te topics will be strengthened over the coming years.
LOCATION: Cambridge University Engineering Department\,  Seminar Room JDB
END:VEVENT
END:VCALENDAR