BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//jEvents 2.0 for Joomla//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:bmh2f88gjtsavb7uon7hd692fc@google.com
CATEGORIES:RTportal.ru
CREATED:20170606T092329
SUMMARY:Структурные модели и глубинное обучение: Acceleration of Convolutional Neural Networks through Elimination of Redundant Convolutions
LOCATION:ИППИ, 6 этаж, 615 аудитория
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:Очередное заседание семинара "Структурные модели и глубинное обучение" сост
 оится 21 марта (понедельник), в 18.30, ИППИ, 6 этаж, 615 аудитория.<br/>Охр
 анник на проходной будет предупрежден, проблем с проходом не будет.<br/><br
 />Название: Acceleration of Convolutional Neural Networks through Eliminati
 on of Redundant Convolutions<br/><br/>Докладчик: Михаил Фигурнов<br/><br/>А
 ннотация: We propose a novel approach to reduce the computational cost of e
 valuation of convolutional neural networks, a factor that has hindered thei
 r deployment in low-power devices such as mobile phones. Inspired by the lo
 op perforation technique from source code optimization, we speed up the bot
 tleneck convolutional layers by skipping their evaluation in some of the sp
 atial positions. We propose and analyze several strategies of choosing thes
 e positions. Our method allows to reduce the evaluation time of modern conv
 olutional neural networks by 50% with a small decrease in accuracy. This ev
 ent was imported from: https://rtportal.ru/index.php/kalendar-sobytij/event
 detail/105/-/strukturnye-modeli-i-glubinnoe-obuchenie-acceleration-of-convo
 lutional-neural-networks-through-elimination-of-redundant-convolutions?tmpl
 =component
DTSTAMP:20260812T201355Z
DTSTART;TZID=Europe/Moscow:20160321T183000
DTEND;TZID=Europe/Moscow:20160321T200000
SEQUENCE:0
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR