{"id":"https://openalex.org/W4415273203","doi":"https://doi.org/10.1145/3748522.3779753","title":"GranQ: Efficient Channel-wise Quantization via Vectorized Pre-Scaling for Zero-Shot QAT","display_name":"GranQ: Efficient Channel-wise Quantization via Vectorized Pre-Scaling for Zero-Shot QAT","publication_year":2025,"publication_date":"2025-03-24","ids":{"openalex":"https://openalex.org/W4415273203","doi":"https://doi.org/10.1145/3748522.3779753"},"language":"en","primary_location":{"id":"doi:10.1145/3748522.3779753","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3748522.3779753","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 41st ACM/SIGAPP Symposium on Applied Computing","raw_type":"proceedings-article"},"type":"article","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3748522.3779753","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Hong, Inpyo","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Hong, Inpyo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101648075","display_name":"Youngwan Jo","orcid":"https://orcid.org/0000-0003-2773-7670"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jo, Youngwan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089029018","display_name":"Hyojeong Lee","orcid":"https://orcid.org/0000-0002-2046-3091"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Hyojeong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102022799","display_name":"Sunghyun Ahn","orcid":"https://orcid.org/0000-0001-9881-5731"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ahn, Sunghyun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Lee, Kijung","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Kijung","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100322261","display_name":"Sanghyun Park","orcid":"https://orcid.org/0000-0001-8564-2632"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Sanghyun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.38123525,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1455","last_page":"1462"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13114","display_name":"Image Processing Techniques and Applications","score":0.9934999942779541,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T13114","display_name":"Image Processing Techniques and Applications","score":0.9934999942779541,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10500","display_name":"Sparse and Compressive Sensing Techniques","score":0.9933000206947327,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10271","display_name":"Seismic Imaging and Inversion Techniques","score":0.9901000261306763,"subfield":{"id":"https://openalex.org/subfields/1908","display_name":"Geophysics"},"field":{"id":"https://openalex.org/fields/19","display_name":"Earth and Planetary Sciences"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.8331000208854675},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.656000018119812},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.4050000011920929},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.3937000036239624},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.36959999799728394},{"id":"https://openalex.org/keywords/deep-neural-networks","display_name":"Deep neural networks","score":0.3635999858379364}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8331000208854675},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7333999872207642},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.656000018119812},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5166000127792358},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4050000011920929},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.3937000036239624},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.36959999799728394},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.3635999858379364},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.33489999175071716},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.32850000262260437},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.32850000262260437},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.30630001425743103},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.26739999651908875},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2630999982357025}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1145/3748522.3779753","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3748522.3779753","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 41st ACM/SIGAPP Symposium on Applied Computing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2503.18339","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.18339","pdf_url":"https://arxiv.org/pdf/2503.18339","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2503.18339","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2503.18339","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.1145/3748522.3779753","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3748522.3779753","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 41st ACM/SIGAPP Symposium on Applied Computing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Zero-shot":[0],"quantization":[1,88,130,140],"(ZSQ)":[2],"enables":[3,125],"neural":[4],"network":[5],"compression":[6],"without":[7],"original":[8],"training":[9,146],"data,":[10,27],"making":[11],"it":[12],"a":[13,85,112],"promising":[14],"solution":[15],"for":[16,23],"restricted":[17],"data":[18],"access":[19],"scenarios.":[20],"To":[21,54,77],"compensate":[22],"the":[24,38,69,74,172,180],"lack":[25],"of":[26],"recent":[28],"ZSQ":[29,155],"methods":[30,58,99,156],"typically":[31,59],"rely":[32],"on":[33,175,183],"synthetic":[34,43],"inputs":[35,44],"generated":[36],"from":[37],"full-precision":[39,181],"model.":[40],"However,":[41],"these":[42],"often":[45],"lead":[46],"to":[47,68,127,167],"activation":[48,87],"distortion,":[49],"especially":[50],"under":[51,144],"low-bit":[52,139],"settings.":[53,141],"mitigate":[55],"this,":[56],"existing":[57],"employ":[60],"per-channel":[61],"scaling,":[62],"but":[63],"they":[64],"still":[65],"struggle":[66],"due":[67],"severe":[70],"computational":[71,135],"overhead":[72],"during":[73,105],"accumulation":[75],"process.":[76],"overcome":[78],"this":[79],"critical":[80],"bottleneck,":[81],"we":[82],"propose":[83],"GranQ,":[84],"novel":[86],"framework":[89],"that":[90,100,150],"introduces":[91],"an":[92],"efficient":[93],"pre-scaling":[94,113],"strategy.":[95],"Unlike":[96],"conventional":[97],"channel-wise":[98],"repeatedly":[101],"perform":[102],"scaling":[103,109,121],"operations":[104],"accumulation,":[106],"GranQ":[107,126,151],"applies":[108],"factors":[110],"in":[111,138,171],"step":[114],"through":[115],"fully":[116],"vectorized":[117],"computation,":[118],"eliminating":[119],"runtime":[120],"overhead.":[122],"This":[123],"design":[124],"maintain":[128],"fine-grained":[129],"accuracy":[131,170],"while":[132],"significantly":[133],"reducing":[134],"burden,":[136],"particularly":[137],"Extensive":[142],"experiments":[143],"quantization-aware":[145],"(QAT)":[147],"settings":[148],"demonstrate":[149],"consistently":[152],"outperforms":[153],"state-of-the-art":[154],"across":[157],"CIFAR":[158],"and":[159,177],"ImageNet.":[160],"In":[161],"particular,":[162],"our":[163],"method":[164],"achieves":[165],"up":[166],"5.45%":[168],"higher":[169],"3-bit":[173],"setting":[174],"CIFAR-100":[176],"even":[178],"surpasses":[179],"baseline":[182],"CIFAR-10.":[184]},"counts_by_year":[],"updated_date":"2026-06-10T14:10:52.464848","created_date":"2025-10-17T00:00:00"}