{"id":"https://openalex.org/W4380353899","doi":"https://doi.org/10.48550/arxiv.2306.05785","title":"End-to-End Neural Network Compression via $\\frac{\\ell_1}{\\ell_2}$ Regularized Latency Surrogates","display_name":"End-to-End Neural Network Compression via $\\frac{\\ell_1}{\\ell_2}$ Regularized Latency Surrogates","publication_year":2023,"publication_date":"2023-06-09","ids":{"openalex":"https://openalex.org/W4380353899","doi":"https://doi.org/10.48550/arxiv.2306.05785"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2306.05785","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2306.05785","pdf_url":"https://arxiv.org/pdf/2306.05785","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2306.05785","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5058454079","display_name":"Anshul Nasery","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nasery, Anshul","raw_affiliation_strings":["Google Research India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research India","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Shah, Hardik","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shah, Hardik","raw_affiliation_strings":["Google Research India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research India","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078446727","display_name":"Arun Sai Suggala","orcid":"https://orcid.org/0000-0003-4113-5924","observed_orcids":["https://orcid.org/0000-0003-4113-5924"]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Suggala, Arun Sai","raw_affiliation_strings":["Google Research India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research India","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101633144","display_name":"Prateek Jain","orcid":"https://orcid.org/0000-0002-9233-5160","observed_orcids":["https://orcid.org/0000-0002-9233-5160"]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jain, Prateek","raw_affiliation_strings":["Google Research India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research India","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1291425158"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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":0.9994999766349792,"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":0.9994999766349792,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.0003000000142492354,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":9.999999747378752e-05,"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/neural-network-compression","display_name":"neural network compression","score":0.9909999966621399},{"id":"https://openalex.org/keywords/l1-and-l2-regularization","display_name":"L1 and L2 regularization","score":0.8009999990463257},{"id":"https://openalex.org/keywords/neural-architecture-search","display_name":"neural architecture search","score":0.7009999752044678},{"id":"https://openalex.org/keywords/pruning","display_name":"pruning","score":0.5619999766349792},{"id":"https://openalex.org/keywords/quantization","display_name":"quantization","score":0.9800000190734863},{"id":"https://openalex.org/keywords/low-rank-matrix-factorization","display_name":"low-rank factorization","score":0.996999979019165},{"id":"https://openalex.org/keywords/bert","display_name":"BERT","score":0.7490000128746033},{"id":"https://openalex.org/keywords/mobilenetv3","display_name":"MobileNetV3","score":0.9980000257492065},{"id":"https://openalex.org/keywords/efficientnet","display_name":"EfficientNet","score":0.949999988079071},{"id":"https://openalex.org/keywords/transfer-learning","display_name":"transfer learning","score":0.6859999895095825},{"id":"https://openalex.org/keywords/training-cost-reduction","display_name":"training cost reduction","score":0.4099999964237213}],"concepts":[{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.8897786736488342},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5919941663742065},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5742970705032349},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5042775869369507},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4867808520793915},{"id":"https://openalex.org/C133095886","wikidata":"https://www.wikidata.org/wiki/Q1307173","display_name":"Single-precision floating-point format","level":3,"score":0.47832179069519043},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4767507016658783},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.4549732208251953},{"id":"https://openalex.org/C187834632","wikidata":"https://www.wikidata.org/wiki/Q188804","display_name":"Factorization","level":2,"score":0.44595789909362793},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4356396198272705},{"id":"https://openalex.org/C25797200","wikidata":"https://www.wikidata.org/wiki/Q828137","display_name":"Compression ratio","level":3,"score":0.43372124433517456},{"id":"https://openalex.org/C56951928","wikidata":"https://www.wikidata.org/wiki/Q3539213","display_name":"Trimming","level":2,"score":0.4179195463657379},{"id":"https://openalex.org/C84211073","wikidata":"https://www.wikidata.org/wiki/Q117879","display_name":"Floating point","level":2,"score":0.4068525433540344},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3556203842163086},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.22305595874786377},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.101938396692276},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0},{"id":"https://openalex.org/C511840579","wikidata":"https://www.wikidata.org/wiki/Q12757","display_name":"Internal combustion engine","level":2,"score":0.0},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.0},{"id":"https://openalex.org/C171146098","wikidata":"https://www.wikidata.org/wiki/Q124192","display_name":"Automotive engineering","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2306.05785","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2306.05785","pdf_url":"https://arxiv.org/pdf/2306.05785","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2306.05785","doi":"https://doi.org/10.48550/arxiv.2306.05785","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2306.05785","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2306.05785","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2306.05785","pdf_url":"https://arxiv.org/pdf/2306.05785","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4380353899.pdf","grobid_xml":"https://content.openalex.org/works/W4380353899.grobid-xml"},"referenced_works_count":26,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2242818861","https://openalex.org/W2553303224","https://openalex.org/W2707890836","https://openalex.org/W2771655537","https://openalex.org/W2805003733","https://openalex.org/W2810075754","https://openalex.org/W2915589364","https://openalex.org/W2950967261","https://openalex.org/W2962750597","https://openalex.org/W2969515962","https://openalex.org/W2970562249","https://openalex.org/W2979691890","https://openalex.org/W2994985593","https://openalex.org/W2995463996","https://openalex.org/W2995738369","https://openalex.org/W3001279689","https://openalex.org/W3036979418","https://openalex.org/W3037932933","https://openalex.org/W3044604993","https://openalex.org/W3125682525","https://openalex.org/W3170437634","https://openalex.org/W3207460419","https://openalex.org/W4200630931","https://openalex.org/W4319166707","https://openalex.org/W7133186778"],"related_works":["https://openalex.org/W4402916496","https://openalex.org/W3212947578","https://openalex.org/W2893550440","https://openalex.org/W7152095762","https://openalex.org/W4319302538","https://openalex.org/W2119144962","https://openalex.org/W3193556940","https://openalex.org/W3082934877","https://openalex.org/W4318751542","https://openalex.org/W2733579387"],"abstract_inverted_index":{"Neural":[0],"network":[1],"(NN)":[2],"compression":[3,12,78,115,162],"via":[4,30,61],"techniques":[5],"such":[6],"as":[7,99],"pruning,":[8,81],"quantization":[9],"requires":[10],"setting":[11],"hyperparameters":[13],"(e.g.,":[14],"number":[15],"of":[16,97,182],"channels":[17],"to":[18],"be":[19,37,73],"pruned,":[20],"bitwidths":[21],"for":[22,24,51,58,165],"quantization)":[23],"each":[25],"layer":[26],"either":[27],"manually":[28],"or":[29,57],"neural":[31],"architecture":[32],"search":[33],"(NAS)":[34],"which":[35,103],"can":[36,72],"computationally":[38],"expensive.":[39],"We":[40],"address":[41],"this":[42],"problem":[43],"by":[44],"providing":[45],"an":[46],"end-to-end":[47],"technique":[48,172],"that":[49],"optimizes":[50],"model's":[52],"Floating":[53],"Point":[54],"Operations":[55],"(FLOPs)":[56],"on-device":[59,147],"latency":[60,65,148],"a":[62,105],"novel":[63],"$\\frac{\\ell_1}{\\ell_2}$":[64],"surrogate.":[66],"Our":[67],"algorithm":[68],"is":[69,88,104],"versatile":[70],"and":[71,84,90,143,190],"used":[74],"with":[75,126],"many":[76],"popular":[77],"methods":[79],"including":[80],"low-rank":[82],"factorization,":[83],"quantization.":[85],"Crucially,":[86],"it":[87],"fast":[89],"runs":[91],"in":[92,124,130,141,146,151],"almost":[93,185],"the":[94,186],"same":[95,187],"amount":[96],"time":[98],"single":[100],"model":[101],"training;":[102],"significant":[106],"training":[107,158,188],"speed-up":[108],"over":[109],"standard":[110,178],"NAS":[111],"methods.":[112],"For":[113,132],"BERT":[114],"on":[116,135,168],"GLUE":[117],"fine-tuning":[118],"tasks,":[119],"we":[120,137],"achieve":[121,138],"$50\\%$":[122],"reduction":[123,140,145],"FLOPs":[125],"only":[127],"$1\\%$":[128],"drop":[129,150],"performance.":[131],"compressing":[133],"MobileNetV3":[134],"ImageNet-1K,":[136],"$15\\%$":[139],"FLOPs,":[142],"$11\\%$":[144],"without":[149],"accuracy,":[152],"while":[153],"still":[154],"requiring":[155],"$3\\times$":[156],"less":[157],"compute":[159],"than":[160,177],"SOTA":[161],"techniques.":[163],"Finally,":[164],"transfer":[166],"learning":[167],"smaller":[169],"datasets,":[170],"our":[171],"identifies":[173],"$1.2\\times$-$1.4\\times$":[174],"cheaper":[175],"architectures":[176,183],"MobileNetV3,":[179],"EfficientNet":[180],"suite":[181],"at":[184],"cost":[189],"accuracy.":[191]},"counts_by_year":[{"year":2024,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2023-06-13T00:00:00"}
