{"id":"https://openalex.org/W3200938796","doi":"https://doi.org/10.1109/ijcnn52387.2021.9533724","title":"Hybrid and non-uniform quantization methods using retro synthesis data for efficient inference","display_name":"Hybrid and non-uniform quantization methods using retro synthesis data for efficient inference","publication_year":2021,"publication_date":"2021-07-18","ids":{"openalex":"https://openalex.org/W3200938796","doi":"https://doi.org/10.1109/ijcnn52387.2021.9533724","mag":"3200938796"},"language":"en","primary_location":{"id":"doi:10.1109/ijcnn52387.2021.9533724","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn52387.2021.9533724","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5001988102","display_name":"GVSL Tej Pratap","orcid":null},"institutions":[{"id":"https://openalex.org/I4210139030","display_name":"Samsung (India)","ror":"https://ror.org/04cpx2569","country_code":"IN","type":"company","lineage":["https://openalex.org/I2250650973","https://openalex.org/I4210139030"]}],"countries":["IN"],"is_corresponding":true,"raw_author_name":"GVSL Tej Pratap","raw_affiliation_strings":["On-Device AI, Samsung Research Institute, Bangalore, India"],"affiliations":[{"raw_affiliation_string":"On-Device AI, Samsung Research Institute, Bangalore, India","institution_ids":["https://openalex.org/I4210139030"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085131561","display_name":"R. Hendra Kumar","orcid":null},"institutions":[{"id":"https://openalex.org/I4210139030","display_name":"Samsung (India)","ror":"https://ror.org/04cpx2569","country_code":"IN","type":"company","lineage":["https://openalex.org/I2250650973","https://openalex.org/I4210139030"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Raja Kumar","raw_affiliation_strings":["On-Device AI, Samsung Research Institute, Bangalore, India"],"affiliations":[{"raw_affiliation_string":"On-Device AI, Samsung Research Institute, Bangalore, India","institution_ids":["https://openalex.org/I4210139030"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5082096064","display_name":"Pradeep NS","orcid":null},"institutions":[{"id":"https://openalex.org/I4210139030","display_name":"Samsung (India)","ror":"https://ror.org/04cpx2569","country_code":"IN","type":"company","lineage":["https://openalex.org/I2250650973","https://openalex.org/I4210139030"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"NS Pradeep","raw_affiliation_strings":["On-Device AI, Samsung Research Institute, Bangalore, India"],"affiliations":[{"raw_affiliation_string":"On-Device AI, Samsung Research Institute, Bangalore, India","institution_ids":["https://openalex.org/I4210139030"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5001988102"],"corresponding_institution_ids":["https://openalex.org/I4210139030"],"apc_list":null,"apc_paid":null,"fwci":0.0961,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.3975817,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":94},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9939000010490417,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9939000010490417,"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/T11698","display_name":"Underwater Acoustics Research","score":0.9818999767303467,"subfield":{"id":"https://openalex.org/subfields/1910","display_name":"Oceanography"},"field":{"id":"https://openalex.org/fields/19","display_name":"Earth and Planetary Sciences"},"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":0.9803000092506409,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6793675422668457},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.6701678037643433},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6505062580108643},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3761349320411682},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.33989614248275757}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6793675422668457},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6701678037643433},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6505062580108643},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3761349320411682},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.33989614248275757}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/ijcnn52387.2021.9533724","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn52387.2021.9533724","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":34,"referenced_works":["https://openalex.org/W1902934009","https://openalex.org/W2108598243","https://openalex.org/W2163605009","https://openalex.org/W2194775991","https://openalex.org/W2233116163","https://openalex.org/W2300242332","https://openalex.org/W2506786467","https://openalex.org/W2560017826","https://openalex.org/W2809624076","https://openalex.org/W2884150179","https://openalex.org/W2915478146","https://openalex.org/W2962298324","https://openalex.org/W2963114950","https://openalex.org/W2963122961","https://openalex.org/W2963163009","https://openalex.org/W2964228333","https://openalex.org/W2970601456","https://openalex.org/W2976783886","https://openalex.org/W2981751377","https://openalex.org/W2982041622","https://openalex.org/W2998218113","https://openalex.org/W3034351824","https://openalex.org/W3034940165","https://openalex.org/W6639703010","https://openalex.org/W6684191040","https://openalex.org/W6698200048","https://openalex.org/W6730047919","https://openalex.org/W6734062232","https://openalex.org/W6753069482","https://openalex.org/W6753469380","https://openalex.org/W6759144272","https://openalex.org/W6767298317","https://openalex.org/W6768457659","https://openalex.org/W6779872307"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W2051487156","https://openalex.org/W2073681303","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Existing":[0],"quantization":[1,7,13,46,121,171,213],"aware":[2],"training":[3,18,36,53],"methods":[4,8,22,85,122,237],"and":[5,75,92,97,104,110,126,144,183,203,221],"post-training":[6,45,120],"attempt":[9],"to":[10,65,153,169],"compensate":[11],"the":[12,50,70,134,162,187,191,196,217,225,239],"loss":[14],"by":[15,58],"leveraging":[16],"on":[17,94,108,186,216,238],"data.":[19,37,54],"Hence,":[20],"these":[21,204],"are":[23,32,151,198,206],"not":[24,88],"effective":[25],"for":[26,52,79,101,139,181],"privacy":[27],"constraint":[28],"applications":[29],"as":[30,66,167],"they":[31],"tightly":[33],"coupled":[34],"with":[35,96,208],"In":[38,190],"contrast,":[39],"this":[40,174],"paper":[41],"proposes":[42],"a":[43,60,209],"data-independent":[44],"scheme":[47,132],"that":[48,150],"eliminates":[49],"need":[51],"This":[55,81,228],"is":[56],"achieved":[57],"generating":[59],"faux":[61],"dataset,":[62],"hereafter":[63],"referred":[64],"\u2018Retro-Synthesis":[67],"Data\u2019,":[68],"from":[69],"FP32":[71,177],"model":[72],"layer":[73,138],"statistics":[74],"further":[76],"using":[77],"it":[78],"quantization.":[80],"approach":[82],"outperformed":[83,176],"state-of-the-art":[84,236],"including,":[86],"but":[87],"limited":[89],"to,":[90],"ZeroQ":[91],"DFQ":[93],"models":[95,149,185],"without":[98],"Batch-Normalization":[99],"layers":[100],"8,":[102],"6,":[103],"4":[105],"bit":[106],"precisions":[107],"ImageNet":[109,188,240],"CIFAR-10":[111],"datasets.":[112],"We":[113],"also":[114],"introduced":[115],"two":[116],"futuristic":[117],"variants":[118],"of":[119,136,212,219],"namely":[123],"\u2018Hybrid":[124],"Quantization\u2019":[125],"\u2018Non-Uniform":[127],"Quantization\u2019.":[128],"The":[129],"Hybrid":[130],"Quantization":[131,194],"determines":[133],"sensitivity":[135],"each":[137],"per-tensor":[140],"&":[141],"per-channel":[142,170],"quantization,":[143],"thereby":[145],"generates":[146],"hybrid":[147],"quantized":[148],"\u201810":[152],"20%\u2019":[154],"more":[155],"efficient":[156],"in":[157,224,231],"inference":[158],"time":[159],"while":[160],"achieving":[161],"same":[163],"or":[164],"better":[165],"accuracy":[166,178,233],"compared":[168],"scheme.":[172],"Also,":[173],"method":[175,229],"when":[179],"applied":[180],"ResNet-18,":[182],"ResNet-SO":[184],"dataset.":[189,241],"proposed":[192],"Non-Uniform":[193],"scheme,":[195],"weights":[197,220],"grouped":[199],"into":[200],"different":[201],"clusters":[202,205],"assigned":[207],"varied":[210],"number":[211,218],"steps":[214],"depending":[215],"their":[222],"ranges":[223],"respective":[226],"cluster.":[227],"resulted":[230],"\u20181%\u2019":[232],"improvement":[234],"against":[235]},"counts_by_year":[{"year":2022,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}