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:20211208T164507Z
LOCATION:3014
DTSTART;TZID=America/Los_Angeles:20211208T103000
DTEND;TZID=America/Los_Angeles:20211208T120000
UID:dac_DAC 2021_sess156@linklings.com
SUMMARY:An Approximate World from Neurons to Genomes
DESCRIPTION:Research Manuscript\n\nThe session features recent advancement
 s in applying approximate computing techniques to different computing para
 digms pertaining to Artificial Intelligence and Machine Learning, includin
 g hyperdimensional systems, spiking neural networks, probabilistic program
 ming, and compressed video analytics. The techniques proposed include new 
 encoding schemes, fixed-point quantization approaches, training methodolog
 ies, and compiler frameworks. The techniques have demonstrated significant
  improvement in both energy consumption and performance across various end
 -applications.\n\nStatheros: Compiler for Efficient Low-Precision Probabil
 istic Programming\n\nLaurel\n\nAs Edge and IoT computing devices process n
 oisy data or make decisions in uncertain environments, they require framew
 orks for inexpensive, yet accurate probabilistic inference. \n\nWe present
  Statheros, the first compiler for low-level, fixed-point approximation of
  probabilistic programming. Statheros...\n\n---------------------\nExact N
 eural Networks from Inexact Multipliers via Fibonacci Weight Encoding\n\nS
 imon\n\nEdge devices must support computationally demanding algorithms, su
 ch as neural networks, within tight area/energy budgets. While approximate
  computing is demonstrated to alleviate these constraints, limiting induce
 d errors remains an open challenge. In this paper, we propose a hardware/s
 oftware co-d...\n\n---------------------\nTCL: an ANN-to-SNN Conversion wi
 th Trainable Clipping Layers\n\nHo\n\nSpiking-neural-networks (SNNs) are p
 romising at edge devices since the event-driven operations of SNNs provide
 s significantly lower power compared to analog-neural-networks (ANNs). Alt
 hough it is difficult to efficiently train SNNs, many techniques to conver
 t trained ANNs to SNNs have been developed...\n\n---------------------\nMO
 VED TO VIRTUAL: Scaling up HBM Efficiency of Top-K SpMV for Approximate Em
 bedding Similarity on FPGAs\n\nParravicini\n\nTop-K SpMV is a key componen
 t of similarity-search on sparse embeddings. This sparse workload does not
  perform well on general-purpose NUMA systems that employ traditional cach
 ing strategies.\nInstead, modern FPGA accelerator cards have a few tricks 
 up their sleeve. We introduce a Top-K SpMV FPGA des...\n\n\nTopic: Design\
 n\nKeyword: Approximate Computing for AI/ML\n\nVirtual Program: Presented 
 In-Person
END:VEVENT
END:VCALENDAR