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:20260402T024533Z
LOCATION:3000\, Level 3
DTSTART;TZID=America/Los_Angeles:20250624T140000
DTEND;TZID=America/Los_Angeles:20250624T141500
UID:dac_DAC 2025_sess101_RESEARCH1242@linklings.com
SUMMARY:NN-AdderNet: Nonnegative and Sparse Weight Optimization for Ultra-
 Low Bitwidth AdderNet Quantization and Compression
DESCRIPTION:Yunxiang Zhang (Binghamton University); Gengchen Sun, Lizhi Fa
 ng, and Biao Sun (Tianjin University); and Wenfeng Zhao (Binghamton Univer
 sity)\n\nEmerging efficient deep neural network (DNN) models, such as Adde
 rNet, have shown great promise in significantly improving hardware efficie
 ncy compared to traditional convolutional neural networks (CNNs). However,
  achieving low bit-width quantization and effective model compression rema
 ins a major challenge. To this end, we introduce Nonnegative AdderNet (NN-
 AdderNet), a quantization- and compression-friendly AdderNet variant that 
 enables model compression down to 4 bits or even lower. We begin by propos
 ing an equivalent transformation of the sum-of-absolute-difference (SAD) k
 ernel in AdderNet, which allows for the formulation of nonnegative weights
 . This transformation effectively eliminates the need for a sign bit, thus
  saving 1 bit per weight. Next, we propose to exploit the dual-sparsity pa
 ttern in the weights of the activation-oriented NN-AdderNet quantized mode
 l. This inherent sparsity enhances the lossless compression performance ov
 er NN-AdderNet. Experimental results show that NN-AdderNet can achieve an 
 average compressed weight bitwidth down to 4 bits or even lower, while ach
 ieving negligible accuracy loss as compared to full-precision AdderNet mod
 els. Such benefits are further illustrated with hardware-level energy and 
 latency improvements in designing DNN inference accelerators. Consequently
 , the NN-AdderNet model exhibits both algorithmic and hardware efficiency,
  thus making it a promising candidate for resource-limited applications.\n
 \nTopics: AI\n\nTracks: AI1: AI/ML Algorithms\n\nSession Chairs: Fan Yang 
 (Fudan University) and Li Shang (Fudan University)\n\n
END:VEVENT
END:VCALENDAR