BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sampling as Optimization - Eric Nalisnick\, University of Cambridg
 e
DTSTART:20190403T124500Z
DTEND:20190403T141500Z
UID:TALK122527@talks.cam.ac.uk
CONTACT:Robert Pinsler
DESCRIPTION:Sampling and optimization are often thought of as alternative 
 methods for model fitting.  In this meeting of the reading group\, we summ
 arize recent results that draw connections between sampling and optimizati
 on.  The key result is the work of Jordan et al. (1998)\, which shows that
  the gradient flow of the Kullback-Leibler divergence in the space of meas
 ures follows the Fokker-Planck equation.  This Fokker-Planck equation can 
 then be recast as running Langevin dynamics in the space of model paramete
 rs.  With this result established\, we then discuss implications for discr
 etization schemes\, viewing SGD as approximate Bayesian inference\, and mo
 dels for which sampling can be faster than optimization.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR