BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211208T164410Z
LOCATION:3014
DTSTART;TZID=America/Los_Angeles:20211208T105000
DTEND;TZID=America/Los_Angeles:20211208T111000
UID:dac_DAC 2021_sess156_RESEARCH423@linklings.com
SUMMARY:TCL: an ANN-to-SNN Conversion with Trainable Clipping Layers
DESCRIPTION:Research Manuscript\n\nTCL: an ANN-to-SNN Conversion with Trai
 nable Clipping Layers\n\nHo\n\nSpiking-neural-networks (SNNs) are promisin
 g at edge devices since the event-driven operations of SNNs provides signi
 ficantly lower power compared to analog-neural-networks (ANNs). Although i
 t is difficult to efficiently train SNNs, many techniques to convert train
 ed ANNs to SNNs have been developed. However, after the conversion, a trad
 e-off relation between accuracy and latency exists in SNNs, causing consid
 erable latency in large size datasets such as ImageNet. We present a techn
 ique, named as TCL, to alleviate the trade-off problem, enabling the accur
 acy of 73.87% (VGG-16) and 70.37% (ResNet-34) for ImageNet with the modera
 te latency of 250 cycles in SNNs.\n\nTopic: Design\n\nKeyword: Approximate
  Computing for AI/ML\n\nVirtual Program: Presented In-Person
END:VEVENT
END:VCALENDAR