{"id":"https://openalex.org/W3201951111","doi":"https://doi.org/10.1109/pact52795.2021.00019","title":"Google Neural Network Models for Edge Devices: Analyzing and Mitigating Machine Learning Inference Bottlenecks","display_name":"Google Neural Network Models for Edge Devices: Analyzing and Mitigating Machine Learning Inference Bottlenecks","publication_year":2021,"publication_date":"2021-09-01","ids":{"openalex":"https://openalex.org/W3201951111","doi":"https://doi.org/10.1109/pact52795.2021.00019","mag":"3201951111"},"language":"en","primary_location":{"id":"doi:10.1109/pact52795.2021.00019","doi":"https://doi.org/10.1109/pact52795.2021.00019","is_oa":false,"landing_page_url":"https://doi.org/10.1109/pact52795.2021.00019","pdf_url":null,"source":{"id":"https://openalex.org/S4210194525","display_name":"International Conference on Parallel Architectures and Compilation Techniques (PACT)","issn_l":"1089-795X","issn":["1089-795X","2641-7944"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 30th International Conference on Parallel Architectures and Compilation Techniques (PACT)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2109.14320","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5016220809","display_name":"Amirali Boroumand","orcid":"https://orcid.org/0000-0001-9748-845X","observed_orcids":["https://orcid.org/0000-0001-9748-845X"]},"institutions":[{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]},{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Amirali Boroumand","raw_affiliation_strings":["Carnegie Mellon Univ","Stanford Univ","Carnegie Mellon University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon Univ","institution_ids":["https://openalex.org/I74973139"]},{"raw_affiliation_string":"Stanford Univ","institution_ids":["https://openalex.org/I97018004"]},{"raw_affiliation_string":"Carnegie Mellon University","institution_ids":["https://openalex.org/I74973139"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036666743","display_name":"Saugata Ghose","orcid":"https://orcid.org/0000-0002-9138-0613","observed_orcids":["https://orcid.org/0000-0002-9138-0613"]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Saugata Ghose","raw_affiliation_strings":["Univ. of Illinois Urbana-Champaign","University of Illinois Urbana Champaign >  >  >  >"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Univ. of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]},{"raw_affiliation_string":"University of Illinois Urbana Champaign >  >  >  >","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054097137","display_name":"Berkin Akin","orcid":"https://orcid.org/0000-0001-6908-5581","observed_orcids":["https://orcid.org/0000-0001-6908-5581"]},"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":"Berkin Akin","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003609686","display_name":"Ravi Narayanaswami","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":"Ravi Narayanaswami","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091016237","display_name":"Geraldo Francisco de Oliveira Junior","orcid":"https://orcid.org/0000-0003-1557-4819","observed_orcids":["https://orcid.org/0000-0003-1557-4819"]},"institutions":[{"id":"https://openalex.org/I35440088","display_name":"ETH Zurich","ror":"https://ror.org/05a28rw58","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I35440088"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Geraldo F. Oliveira","raw_affiliation_strings":["ETH Z\u00fcrich"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ETH Z\u00fcrich","institution_ids":["https://openalex.org/I35440088"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101599421","display_name":"Xiaoyu Ma","orcid":"https://orcid.org/0000-0002-8666-104X","observed_orcids":["https://orcid.org/0000-0002-8666-104X"]},"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":"Xiaoyu Ma","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Eric Shiu","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":"Eric Shiu","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5050695684","display_name":"Onur Mutlu","orcid":"https://orcid.org/0000-0002-0075-2312","observed_orcids":["https://orcid.org/0000-0002-0075-2312"]},"institutions":[{"id":"https://openalex.org/I35440088","display_name":"ETH Zurich","ror":"https://ror.org/05a28rw58","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I35440088"]},{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]}],"countries":["CH","US"],"is_corresponding":false,"raw_author_name":"Onur Mutlu","raw_affiliation_strings":["Carnegie Mellon Univ","ETH Z\u00fcrich"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon Univ","institution_ids":["https://openalex.org/I74973139"]},{"raw_affiliation_string":"ETH Z\u00fcrich","institution_ids":["https://openalex.org/I35440088"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":8.9044,"has_fulltext":true,"cited_by_count":90,"citation_normalized_percentile":{"value":0.98884214,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":88,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"159","last_page":"172"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.47850000858306885,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.47850000858306885,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.3075999915599823,"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.13050000369548798,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/edge-computing","display_name":"edge computing","score":0.7429999709129333},{"id":"https://openalex.org/keywords/machine-learning-accelerators","display_name":"machine learning accelerators","score":0.8790000081062317},{"id":"https://openalex.org/keywords/neural-network-inference","display_name":"neural network inference","score":0.8999999761581421},{"id":"https://openalex.org/keywords/mensa","display_name":"Mensa","score":0.5260000228881836},{"id":"https://openalex.org/keywords/energy-efficiency","display_name":"energy efficiency","score":0.6359999775886536},{"id":"https://openalex.org/keywords/inference-throughput","display_name":"inference throughput","score":0.4339999854564667}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7931891679763794},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.6804049611091614},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6045069694519043},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.601296603679657},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5907178521156311},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5619067549705505},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.5042563676834106},{"id":"https://openalex.org/C2778456923","wikidata":"https://www.wikidata.org/wiki/Q5337692","display_name":"Edge computing","level":3,"score":0.5015735626220703},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.4922086298465729},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.47028282284736633},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.44725069403648376},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.44179484248161316},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4283527135848999},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.19979426264762878},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.12759941816329956},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.09129342436790466},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.08915194869041443},{"id":"https://openalex.org/C146978453","wikidata":"https://www.wikidata.org/wiki/Q3798668","display_name":"Aerospace engineering","level":1,"score":0.0},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.0},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.0},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1109/pact52795.2021.00019","doi":"https://doi.org/10.1109/pact52795.2021.00019","is_oa":false,"landing_page_url":"https://doi.org/10.1109/pact52795.2021.00019","pdf_url":null,"source":{"id":"https://openalex.org/S4210194525","display_name":"International Conference on Parallel Architectures and Compilation Techniques (PACT)","issn_l":"1089-795X","issn":["1089-795X","2641-7944"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 30th International Conference on Parallel Architectures and Compilation Techniques (PACT)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2109.14320","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2109.14320","pdf_url":"https://arxiv.org/pdf/2109.14320","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.2109.14320","doi":"https://doi.org/10.48550/arxiv.2109.14320","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2109.14320","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:2109.14320","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2109.14320","pdf_url":"https://arxiv.org/pdf/2109.14320","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":[{"id":"https://openalex.org/F4320306087","display_name":"Semiconductor Research Corporation","ror":"https://ror.org/047z4n946"},{"id":"https://openalex.org/F4320316785","display_name":"VMware","ror":"https://ror.org/01mwmh458"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W3201951111.pdf","grobid_xml":"https://content.openalex.org/works/W3201951111.grobid-xml"},"referenced_works_count":102,"referenced_works":["https://openalex.org/W613766092","https://openalex.org/W1485009520","https://openalex.org/W1498436455","https://openalex.org/W1508856789","https://openalex.org/W1514535095","https://openalex.org/W1546771929","https://openalex.org/W1568514080","https://openalex.org/W1686810756","https://openalex.org/W1689711448","https://openalex.org/W1810943226","https://openalex.org/W1828163288","https://openalex.org/W1849277567","https://openalex.org/W1895577753","https://openalex.org/W1905882502","https://openalex.org/W1924770834","https://openalex.org/W1934184906","https://openalex.org/W1934410531","https://openalex.org/W1947481528","https://openalex.org/W1989831790","https://openalex.org/W2005708641","https://openalex.org/W2064675550","https://openalex.org/W2067523571","https://openalex.org/W2097117768","https://openalex.org/W2105102111","https://openalex.org/W2112796928","https://openalex.org/W2116261113","https://openalex.org/W2116435618","https://openalex.org/W2117539524","https://openalex.org/W2130942839","https://openalex.org/W2131218797","https://openalex.org/W2152839228","https://openalex.org/W2154579312","https://openalex.org/W2157331557","https://openalex.org/W2160428323","https://openalex.org/W2163605009","https://openalex.org/W2194775991","https://openalex.org/W2289252105","https://openalex.org/W2373570000","https://openalex.org/W2414912620","https://openalex.org/W2466675884","https://openalex.org/W2473640056","https://openalex.org/W2516141709","https://openalex.org/W2525778437","https://openalex.org/W2545177271","https://openalex.org/W2573587735","https://openalex.org/W2585720638","https://openalex.org/W2591013610","https://openalex.org/W2594492285","https://openalex.org/W2605258629","https://openalex.org/W2605347906","https://openalex.org/W2606722458","https://openalex.org/W2612076670","https://openalex.org/W2612445135","https://openalex.org/W2616915022","https://openalex.org/W2625457103","https://openalex.org/W2626991402","https://openalex.org/W2752813595","https://openalex.org/W2762910930","https://openalex.org/W2770159745","https://openalex.org/W2784041417","https://openalex.org/W2786562507","https://openalex.org/W2789554134","https://openalex.org/W2790925711","https://openalex.org/W2796625795","https://openalex.org/W2883780447","https://openalex.org/W2883929540","https://openalex.org/W2884687835","https://openalex.org/W2888243885","https://openalex.org/W2919115771","https://openalex.org/W2921796727","https://openalex.org/W2931743911","https://openalex.org/W2935331687","https://openalex.org/W2944681516","https://openalex.org/W2945146780","https://openalex.org/W2950635152","https://openalex.org/W2952438573","https://openalex.org/W2952453038","https://openalex.org/W2953212265","https://openalex.org/W2962760690","https://openalex.org/W2962949934","https://openalex.org/W2963125010","https://openalex.org/W2963414781","https://openalex.org/W2963451564","https://openalex.org/W2963510045","https://openalex.org/W2964095005","https://openalex.org/W2964199361","https://openalex.org/W2969359473","https://openalex.org/W2976697034","https://openalex.org/W2979577066","https://openalex.org/W2980104813","https://openalex.org/W2980200167","https://openalex.org/W2986423549","https://openalex.org/W3007227084","https://openalex.org/W3023154643","https://openalex.org/W3043023836","https://openalex.org/W3046503725","https://openalex.org/W3113087954","https://openalex.org/W4249932213","https://openalex.org/W4394670654","https://openalex.org/W6629815555","https://openalex.org/W6760385162","https://openalex.org/W7095449876"],"related_works":["https://openalex.org/W3134295251","https://openalex.org/W4295036606","https://openalex.org/W4206272272","https://openalex.org/W4212896416","https://openalex.org/W4323922577","https://openalex.org/W4408888778","https://openalex.org/W4379031098","https://openalex.org/W3010249414","https://openalex.org/W7000336736","https://openalex.org/W2963947694"],"abstract_inverted_index":{"Emerging":[0],"edge":[1,38,58,75,186,249,289],"computing":[2,39],"platforms":[3],"often":[4],"contain":[5],"machine":[6],"learning":[7],"(ML)":[8],"accelerators":[9,60,188],"that":[10,97,138,253],"can":[11],"accelerate":[12],"inference":[13],"for":[14,214,236,276],"a":[15,67,80,129,176,203,261,314],"wide":[16,81],"range":[17,82],"of":[18,36,66,83,143,151,196,202,206,233,246,255,264,274],"neural":[19],"network":[20],"(NN)":[21],"models.":[22],"These":[23],"models":[24,77,157,208,251,278],"are":[25],"designed":[26],"to":[27,169,199,222,269],"fit":[28],"within":[29,92,163],"the":[30,37,64,98,139,144,148,164,170,200,224,231,237,247,256,303],"limited":[31],"area":[32],"and":[33,87,123,132,158,192,209,239,296,300,306,309],"energy":[34,121,131,294],"constraints":[35],"platforms,":[40],"each":[41,89,93,195,215,234],"targeting":[42],"various":[43],"applications":[44],"(e.g.,":[45],"face":[46],"detection,":[47],"speech":[48],"recognition,":[49],"translation,":[50],"image":[51],"captioning,":[52],"video":[53],"analytics).":[54],"To":[55],"understand":[56],"how":[57],"ML":[59,187],"perform,":[61],"we":[62,172],"characterize":[63],"performance":[65,133],"commercial":[68],"Google":[69,74,248,288],"Edge":[70,99,145,304],"TPU,":[71,305],"using":[72],"24":[73,287],"NN":[76,84,90,156,161,166,207,212,216,250,290],"(which":[78],"span":[79],"model":[85],"types)":[86],"analyzing":[88],"layer":[91,225,238],"model.":[94],"We":[95,174],"find":[96],"TPU":[100,146],"suffers":[101],"from":[102],"three":[103,281],"major":[104],"shortcomings:":[105],"(1)":[106],"it":[107,115],"operates":[108,116],"significantly":[109,117],"below":[110,118],"peak":[111],"computational":[112],"throughput,":[113],"(2)":[114],"its":[119,125],"theoretical":[120],"efficiency,":[122],"(3)":[124],"memory":[126],"system":[127],"is":[128],"large":[130],"bottleneck.":[134],"Our":[135,243],"characterization":[136],"reveals":[137],"one-size-fits-all,":[140],"monolithic":[141],"design":[142,270],"ignores":[147],"high":[149],"degree":[150],"heterogeneity":[152],"both":[153,190,230],"across":[154,159,285],"different":[155,160],"layers":[162,257],"same":[165],"model,":[167],"leading":[168],"shortcomings":[171],"observe.":[173],"propose":[175],"new":[177],"acceleration":[178],"framework":[179],"called":[180],"Mensa.":[181],"Mensa":[182,218,275,292],"incorporates":[183],"multiple":[184],"heterogeneous":[185],"(including":[189],"on-chip":[191],"near-data":[193],"accelerators),":[194],"which":[197,220,266],"caters":[198],"characteristics":[201],"particular":[204],"subset":[205],"layers.":[210],"During":[211],"inference,":[213],"layer,":[217],"decides":[219],"accelerator":[221,235],"schedule":[223],"on,":[226],"taking":[227],"into":[228,260],"account":[229],"optimality":[232],"layer-to-layer":[240],"communication":[241],"costs.":[242],"comprehensive":[244],"analysis":[245],"shows":[252],"all":[254,286],"naturally":[258],"group":[259],"small":[262],"number":[263],"clusters,":[265],"allows":[267],"us":[268],"an":[271],"efficient":[272],"implementation":[273],"these":[277],"with":[279],"only":[280],"specialized":[282],"accelerators.":[283],"Averaged":[284],"models,":[291],"improves":[293],"efficiency":[295],"throughput":[297],"by":[298,307],"3.0x":[299],"3.1x":[301],"over":[302,311],"2.4x":[308],"4.3x":[310],"Eyeriss":[312],"v2,":[313],"state-of-the-art":[315],"accelerator.":[316]},"counts_by_year":[{"year":2026,"cited_by_count":20},{"year":2025,"cited_by_count":18},{"year":2024,"cited_by_count":19},{"year":2023,"cited_by_count":18},{"year":2022,"cited_by_count":14},{"year":2021,"cited_by_count":1}],"updated_date":"2026-11-05T12:25:35.769452","created_date":"2021-10-11T00:00:00"}