BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Math of Deep Learning Seminar: Michael Eickenberg
DTSTART:20211214T160000Z
DTEND:20211214T170000Z
DTSTAMP:20261009T014500Z
UID:indico-event-2763@indico.flatironinstitute.org
DESCRIPTION:Title: A greedy and local perspective on neural network traini
 ng and parallelization\n\nAbstract: In this talk\, Michael will show that 
 meaningful object recognition accuracy can be achieved without end-to-end 
 backpropagation\, using only progressive linearization of object category 
 as training objective. The approach used - greedy and local optimization -
  leads to an immediate application in parallelizing the training of deep n
 etworks.\n\nWill present both of these ideas\, based on the papers https:
 //arxiv.org/abs/1812.11446 https://arxiv.org/abs/1901.08164 https://arxi
 v.org/abs/2106.06401 as well as other recent developments in local parall
 el training.\n\n \n\nIf you would like to attend\, please email crampersa
 d@flatironinstitute.org for the Zoom link.\n\nhttps://indico.flatironinst
 itute.org/event/2763/
URL:https://indico.flatironinstitute.org/event/2763/
END:VEVENT
END:VCALENDAR