BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000711Z
LOCATION:401-402
DTSTART;TZID=America/Denver:20231116T133000
DTEND;TZID=America/Denver:20231116T140000
UID:submissions.supercomputing.org_SC23_sess184_pap428@linklings.com
SUMMARY:Structural Coding: A Low-Cost Scheme to Protect CNNs from Large-Gr
 anularity Memory Faults
DESCRIPTION:Ali Asgari Khoshouyeh (University of British Columbia); Floria
 n Geissler, Seyed Qutub, and Michael Paulitsch (Intel Corporation); and Pr
 ashant Nair and Karthik Pattabiraman (University of British Columbia)\n\nT
 he advent of High Performance Computing has led to the adoption of Convolu
 tional Neural Networks (CNNs) in safety-critical applications such as auto
 nomous vehicles. However, CNNs are vulnerable to DRAM errors corrupting th
 eir parameters, thereby degrading their accuracy. Existing techniques for 
 protecting CNNs from DRAM errors are either expensive or fail to protect f
 rom large-granularity, multi-bit errors, which occur commonly in DRAMs.\n\
 nWe propose a software-implemented coding scheme, Structural Coding (SC) f
 or protecting CNNs from large-granularity memory errors. SC achieves three
  orders of magnitude reduction in Silent Data Corruption (SDC) rates of CN
 Ns compared to no protection. Its average error correction coverage is als
 o higher than other software-techniques to protect CNNs from faults in the
  memory. Further, its average performance, memory, and energy overheads ar
 e respectively 3%, 15.71%, and 4.38%. These overheads are much lower than 
 other software protection techniques.\n\nTag: Accelerators, Artificial Int
 elligence/Machine Learning, Codesign, Fault Handling and Tolerance, Perfor
 mance Measurement, Modeling, and Tools, Post-Moore Computing\n\nRegistrati
 on Category: Tech Program Reg Pass\n\nReproducibility Badges: Artifact Ava
 ilable, Artifact Functional\n\nSession Chair: Lishan Yang (George Mason Un
 iversity (GMU))\n\n
END:VEVENT
END:VCALENDAR