{"id":"https://openalex.org/W3004100696","doi":null,"title":"Near-Lossless Post-Training Quantization of Deep Neural Networks via a Piecewise Linear Approximation","display_name":"Near-Lossless Post-Training Quantization of Deep Neural Networks via a Piecewise Linear Approximation","publication_year":2020,"publication_date":"2020-01-31","ids":{"openalex":"https://openalex.org/W3004100696","mag":"3004100696"},"language":"en","primary_location":{"id":"mag:3004100696","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2002.00104","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2002.00104","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5067626381","display_name":"Jun Fang","orcid":"https://orcid.org/0000-0001-7427-4723","observed_orcids":["https://orcid.org/0000-0001-7427-4723"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jun Fang","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0001-7427-4723","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067008084","display_name":"Ali Reza Shafiee","orcid":"https://orcid.org/0000-0001-7154-9138","observed_orcids":["https://orcid.org/0000-0001-7154-9138"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ali Shafiee","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0001-7154-9138","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049737330","display_name":"Hamzah A. Abdel-Aziz","orcid":"https://orcid.org/0000-0002-8991-0451","observed_orcids":["https://orcid.org/0000-0002-8991-0451"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hamzah Abdel-Aziz","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0002-8991-0451","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065376296","display_name":"David Thorsley","orcid":"https://orcid.org/0000-0001-8625-5095","observed_orcids":["https://orcid.org/0000-0001-8625-5095"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"David Thorsley","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0001-8625-5095","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101886642","display_name":"Georgios Georgiadis","orcid":"https://orcid.org/0000-0003-0147-9167","observed_orcids":["https://orcid.org/0000-0003-0147-9167"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Georgios Georgiadis","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0003-0147-9167","affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5012298536","display_name":"Joseph Hassoun","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Joseph Hassoun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":10,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.00039999998989515007,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10363","display_name":"Low-power high-performance VLSI design","score":0.0003000000142492354,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/post-training-quantization","display_name":"post-training quantization","score":0.9990000128746033},{"id":"https://openalex.org/keywords/deep-neural-network","display_name":"deep neural network","score":0.7210000157356262},{"id":"https://openalex.org/keywords/quantization-error-minimization","display_name":"quantization error minimization","score":0.9129999876022339},{"id":"https://openalex.org/keywords/energy-efficient-deployment","display_name":"energy-efficient deployment","score":0.7300000190734863},{"id":"https://openalex.org/keywords/resource-constrained-devices","display_name":"resource-constrained devices","score":0.8510000109672546},{"id":"https://openalex.org/keywords/image-classification","display_name":"image classification","score":0.703000009059906},{"id":"https://openalex.org/keywords/semantic-segmentation","display_name":"semantic segmentation","score":0.9950000047683716},{"id":"https://openalex.org/keywords/object-detection","display_name":"object detection","score":1.0}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8601516485214233},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.599200963973999},{"id":"https://openalex.org/C164660894","wikidata":"https://www.wikidata.org/wiki/Q2037833","display_name":"Piecewise","level":2,"score":0.5614935159683228},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5304754376411438},{"id":"https://openalex.org/C93372532","wikidata":"https://www.wikidata.org/wiki/Q6552455","display_name":"Linde\u2013Buzo\u2013Gray algorithm","level":3,"score":0.4946708381175995},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.47848767042160034},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4686286747455597},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3372941315174103},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.32849812507629395},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.08111879229545593}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:3004100696","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2002.00104","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null}],"best_oa_location":{"id":"mag:3004100696","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2002.00104","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":5,"referenced_works":["https://openalex.org/W2586654419","https://openalex.org/W2604998962","https://openalex.org/W2799243248","https://openalex.org/W2963114950","https://openalex.org/W2964309882"],"related_works":["https://openalex.org/W3108835732","https://openalex.org/W4220934237","https://openalex.org/W6921997703","https://openalex.org/W2985339713","https://openalex.org/W4312600348","https://openalex.org/W4386076113","https://openalex.org/W3035855442","https://openalex.org/W4405265858","https://openalex.org/W4412404073","https://openalex.org/W4324286739"],"abstract_inverted_index":{"Quantization":[0],"plays":[1],"an":[2,109],"important":[3],"role":[4],"in":[5],"the":[6,30,94,119,126],"energy-efficient":[7],"deployment":[8],"of":[9,112],"deep":[10],"neural":[11,46],"networks":[12,47],"on":[13,145],"resource-limited":[14],"devices.":[15],"Post-training":[16],"quantization":[17,40,74,96,113,127,133],"is":[18],"highly":[19],"desirable":[20],"since":[21],"it":[22,55],"does":[23],"not":[24],"require":[25],"retraining":[26],"or":[27],"access":[28],"to":[29,50,63,77,130],"full":[31],"training":[32],"dataset.":[33],"The":[34],"well-established":[35],"uniform":[36],"scheme":[37,76],"for":[38,81,101],"post-training":[39,132],"achieves":[41,142],"satisfactory":[42],"results":[43,136],"by":[44,124],"converting":[45],"from":[48,57],"full-precision":[49],"8-bit":[51],"fixed-point":[52],"integers.":[53],"However,":[54],"suffers":[56],"significant":[58],"performance":[59,144],"degradation":[60],"when":[61],"quantizing":[62],"lower":[64],"bit-widths.":[65],"In":[66],"this":[67],"paper,":[68],"we":[69],"propose":[70],"a":[71],"piecewise":[72],"linear":[73],"(PWLQ)":[75],"enable":[78],"accurate":[79],"approximation":[80],"tensor":[82],"values":[83],"that":[84,117,138],"have":[85],"bell-shaped":[86],"distributions":[87],"with":[88,104,153],"long":[89],"tails.":[90],"Our":[91],"approach":[92],"breaks":[93],"entire":[95,120],"range":[97,121],"into":[98],"non-overlapping":[99],"regions":[100],"each":[102,105],"tensor,":[103],"region":[106],"being":[107],"assigned":[108],"equal":[110],"number":[111],"levels.":[114],"Optimal":[115],"breakpoints":[116],"divide":[118],"are":[122],"found":[123],"minimizing":[125],"error.":[128],"Compared":[129],"state-of-the-art":[131],"methods,":[134],"experimental":[135],"show":[137],"our":[139],"proposed":[140],"method":[141],"superior":[143],"image":[146],"classification,":[147],"semantic":[148],"segmentation,":[149],"and":[150],"object":[151],"detection":[152],"minor":[154],"overhead.":[155]},"counts_by_year":[{"year":2023,"cited_by_count":2},{"year":2022,"cited_by_count":4},{"year":2021,"cited_by_count":3},{"year":2020,"cited_by_count":1}],"updated_date":"2026-11-03T10:50:19.383937","created_date":"2020-02-07T00:00:00"}