BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//jEvents 2.0 for Joomla//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:qkdm7i1rfi82goiqsro2eu92o0@google.com
CATEGORIES:RTportal.ru
CREATED:20170606T092329
SUMMARY:Семинар СМиГО: Tensorizing Neural Networks
LOCATION:ИППИ РАН в 615 аудитории (6 этаж)
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:Докладчик: Александр Новиков (ВШЭ)<br/>Тема: Tensorizing Neural Networks<br
 /><br/>Аннотация:<br/>Convolutional neural networks excel in image recognit
 ion tasks, but this comes at the cost of high computational and memory comp
 lexity. CNNs require millions of floating point operations to process an im
 age and therefore real-time applications<br/>need powerful CPU or GPU devic
 es. Moreover, these networks contain millions of trainable parameters and c
 onsume hundreds of megabytes of storage and memory bandwidth. Thus, CNNs ar
 e forced to use RAM instead of solely relying on the processor cache – orde
 rs of magnitude more energy efficient memory device – which increases the e
 nergy consumption even more. These reasons restrain the spread of CNNs on m
 obile devices. I will talk about our work on tensor factorization framework
  to compress fully-connected and convolutional layers of CNNs. Another rese
 arch direction (besides compression) is to increase the size of the layers 
 by training them in the compact tensor format to increase the accuracy. <br
 /><br/>For more details see papers <br/><a href="https://papers.nips.cc/pap
 er/5787-tensorizing-neural-networks">https://papers.nips.cc/paper/5787-tens
 orizing-neural-networks</a><br/><a href="https://arxiv.org/abs/1611.03214">
 https://arxiv.org/abs/1611.03214</a> This event was imported from: https://
 rtportal.ru/index.php/kalendar-sobytij/eventdetail/121/-/seminar-smigo-tens
 orizing-neural-networks?tmpl=component
DTSTAMP:20260812T215858Z
DTSTART;TZID=Europe/Moscow:20161122T183000
DTEND;TZID=Europe/Moscow:20161122T200000
SEQUENCE:0
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR