BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160553Z
LOCATION:Track 10
DTSTART;TZID=America/New_York:20201111T153000
DTEND;TZID=America/New_York:20201111T155500
UID:submissions.supercomputing.org_SC20_sess205_ws_corr107@linklings.com
SUMMARY:Correctness-Preserving Compression of Datasets and Neural Network 
 Models
DESCRIPTION:Workshop\n\nCorrectness-Preserving Compression of Datasets and
  Neural Network Models\n\nJoseph, Chalapathi, Bhaskara, Gopalakrishnan, Pa
 nchekha...\n\nNeural networks deployed on edge devices must be efficient b
 oth in terms of their model size and the amount of data movement they caus
 e when classifying inputs. These efficiencies are typically achieved throu
 gh model compression: pruning a fully trained network model by zeroing out
  the weights. Given the overall challenge of neural network correctness, w
 e argue that focusing on correctness preservation may allow the community 
 to make measurable progress. We present a state-of-the-art model compressi
 on framework called Condensa around which we have launched correctness pre
 servation studies. After presenting Condensa, we describe our initial effo
 rts at understanding the effect of model compression in semantic terms, go
 ing beyond the top n% accuracy on which Condensa is currently based. We al
 so take up the relatively unexplored direction of data compression which m
 ay help reduce data movement. We report preliminary results of learning fr
 om decompressed data to understand the effects of compression artifacts. L
 earning without decompressing input data also holds promise in terms of bo
 osting efficiency, and we also report preliminary results in this regard. 
 Our experiments centered around a state-of-the-art model compression frame
 work called Condensa and two data compression algorithms, namely JPEG and 
 ZFP, demonstrate the potential for employing model and dataset compression
  without adversely affecting correctness.\n\nTag: Correctness, Data Analyt
 ics, Compression, and Management, Machine Learning, Deep Learning and Arti
 ficial Intelligence\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR