BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Theoretical guarantees for sampling and inference in generative mo
 dels with latent diffusions - Belinda Tzen -  University of Illinois at Ur
 bana-Champaign
DTSTART:20210629T140000Z
DTEND:20210629T150000Z
UID:TALK161191@talks.cam.ac.uk
CONTACT:Francisco Vargas
DESCRIPTION:*Paper:*\n\nTalk is based on "this":http://proceedings.mlr.pre
 ss/v99/tzen19a.html paper.\n\n*Abstract:*\n\nWe introduce and study a clas
 s of probabilistic generative models\, where the latent object is a finite
 -dimensional diffusion process on a finite time interval and the observed 
 variable is drawn conditionally on the terminal point of the diffusion. We
  make the following contributions: We provide a unified viewpoint on both 
 sampling and variational inference in such generative models through the l
 ens of stochastic control. We quantify the expressiveness of diffusion-bas
 ed generative models. Specifically\, we show that one can efficiently samp
 le from a wide class of terminal target distributions by choosing the drif
 t of the latent diffusion from the class of multilayer feedforward neural 
 nets\, with the accuracy of sampling measured by the Kullback–Leibler di
 vergence to the target distribution. Finally\, we present and analyze a sc
 heme for unbiased simulation of generative models with latent diffusions a
 nd provide bounds on the variance of the resulting estimators. This scheme
  can be implemented as a deep generative model with a random number of lay
 ers.\n\n*Website:* https://scholar.google.com/citations?user=UgdTN9UAAAAJ&
 hl=en\n\nPart of ML@CL Seminar Series in topics relevant to machine learni
 ng and statistics.\n\nMeeting ID: 966 8293 0826\n\nPasscode: 007922
LOCATION:https://cl-cam-ac-uk.zoom.us/j/96682930826?pwd=SVVpTFplRVRFeXhmOE
 1VejFVeTdzdz09
END:VEVENT
END:VCALENDAR