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:20260730T152637Z
LOCATION:
DTSTART;TZID=America/Los_Angeles:20260728T123000
DTEND;TZID=America/Los_Angeles:20260728T123000
UID:dac_DAC 2026_sess315_RESEARCH800@linklings.com
SUMMARY:E³-CODE: Embedded and Efficient Error-Correcting Code for Error-Re
 silient Neural Networks
DESCRIPTION:Venkata Nithin Kamineni, Habibur Rahaman, Ovishake Sen, Baibha
 b Chatterjee, SWARUP BHUNIA, and Rickard Ewetz (University of Florida)\n\n
 Neural networks are widely deployed at the edge to process high-\ndimensio
 nal sensor data, but they are susceptible to burst errors that\ncan corrup
 t weights and degrade inference accuracy. Conventional\nerror-correcting c
 odes (ECC) mitigate errors but incur significant\nmemory overhead. Recent 
 ECC methods for neural networks over-\nwrite the least significant bits of
  the model weights with parity bits,\nproviding zero-overhead resilience a
 t the expense of slightly re-\nduced inference accuracy. In this paper, we
  propose a framework for\nEmbedded and Efficient Error-Correcting Code for
  Error-Resilient\nNeural Networks called (E3-CODE). The proposed method em
 beds\nmulti-bit parity within the entire weight representation, which is\n
 different from only modifying the LSBs of the weights. To mini-\nmize the 
 negative impact from the parity embedded ECC, weight\nand parity assignmen
 ts are jointly optimized via a mixed-integer\nlinear programming (MILP) fo
 rmulation. We also propose a hybrid\nECC scheme that combines the embedded
  ECC with conventional\nECC to trade-off minor memory overhead for signifi
 cantly im-\nproved reliance. The experimental evaluation on the ImageNet a
 nd\nCIFAR-10 datasets using ResNet, MobileNetV2, and EfficientNet-B0\ndemo
 nstrates that E3-CODE maintains software-level accuracy in\nthe presence o
 f burst errors. Compared with prior methods, the\nlifetime of the edge sys
 tem is extended by 4.8𝑋 with no memory\noverhead and 10𝑋 with less than 2%
  memory overhead.\n\nTopics: Quantum\n\nTrack: DES6. Quantum Computing, Sy
 stems\n\nSession Chairs: Hokeun Kim (Arizona State University); BaekGyu Ki
 m (Daegu Gyeongbuk Institute of Science and Technology); Xiang Chen (Pekin
 g University); Sangyeob Kim (Yonsei University); Robert M. Radway (Univers
 ity of Pennsylvania); Prashanth Mohan (University of Pittsburgh); Xiaoxuan
  Yang (University of Virginia, Stanford University); Ramtin Zand (Universi
 ty of South Carolina); Hadi Kamali (University of Central Florida); Jiafen
 g Xie (Villanova University); Prabuddha Chakraborty (University of Maine);
  Junichiro Kadomoto (University of Tokyo); and Yosuke Ueno (RIKEN)\n\n
END:VEVENT
END:VCALENDAR