{"id":"https://openalex.org/W4229055905","doi":"https://doi.org/10.1109/percomworkshops53856.2022.9767442","title":"Enabling On-Device Smartphone GPU based Training: Lessons Learned","display_name":"Enabling On-Device Smartphone GPU based Training: Lessons Learned","publication_year":2022,"publication_date":"2022-03-21","ids":{"openalex":"https://openalex.org/W4229055905","doi":"https://doi.org/10.1109/percomworkshops53856.2022.9767442"},"language":"en","primary_location":{"id":"doi:10.1109/percomworkshops53856.2022.9767442","is_oa":false,"landing_page_url":"https://doi.org/10.1109/percomworkshops53856.2022.9767442","pdf_url":null,"source":{"id":"https://openalex.org/S4363608020","display_name":"2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops)","raw_type":"proceedings-article"},"type":"conference-paper","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/A5101322224","display_name":"Anish Das","orcid":null},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Anish Das","raw_affiliation_strings":["University of Cambridge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Cambridge","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102990915","display_name":"Young D. Kwon","orcid":"https://orcid.org/0000-0002-5216-9057"},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Young D. Kwon","raw_affiliation_strings":["University of Cambridge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Cambridge","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021750511","display_name":"Jagmohan Chauhan","orcid":"https://orcid.org/0000-0003-2080-3276"},"institutions":[{"id":"https://openalex.org/I43439940","display_name":"University of Southampton","ror":"https://ror.org/01ryk1543","country_code":"GB","type":"education","lineage":["https://openalex.org/I43439940"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Jagmohan Chauhan","raw_affiliation_strings":["University of Southampton"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Southampton","institution_ids":["https://openalex.org/I43439940"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5010623957","display_name":"Cecilia Mascolo","orcid":"https://orcid.org/0000-0001-9614-4380"},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Cecilia Mascolo","raw_affiliation_strings":["University of Cambridge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Cambridge","institution_ids":["https://openalex.org/I241749"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":15,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"533","last_page":"538"},"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/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9965000152587891,"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"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9954000115394592,"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/flops","display_name":"FLOPS","score":0.8881000280380249},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.882093071937561},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.8168809413909912},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5662527680397034},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5460427403450012},{"id":"https://openalex.org/keywords/mobile-device","display_name":"Mobile device","score":0.543236255645752},{"id":"https://openalex.org/keywords/mobile-processor","display_name":"Mobile processor","score":0.4402988851070404},{"id":"https://openalex.org/keywords/memory-bandwidth","display_name":"Memory bandwidth","score":0.4238308370113373},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4207311272621155},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.31947147846221924},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.2509133219718933},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.18921524286270142},{"id":"https://openalex.org/keywords/mobile-technology","display_name":"Mobile technology","score":0.1321951150894165},{"id":"https://openalex.org/keywords/mobile-web","display_name":"Mobile Web","score":0.10014212131500244},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.09480488300323486}],"concepts":[{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.8881000280380249},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.882093071937561},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.8168809413909912},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5662527680397034},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5460427403450012},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.543236255645752},{"id":"https://openalex.org/C1665295","wikidata":"https://www.wikidata.org/wiki/Q6887219","display_name":"Mobile processor","level":5,"score":0.4402988851070404},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.4238308370113373},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4207311272621155},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.31947147846221924},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2509133219718933},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.18921524286270142},{"id":"https://openalex.org/C60952562","wikidata":"https://www.wikidata.org/wiki/Q6887246","display_name":"Mobile technology","level":3,"score":0.1321951150894165},{"id":"https://openalex.org/C516764902","wikidata":"https://www.wikidata.org/wiki/Q1043805","display_name":"Mobile Web","level":4,"score":0.10014212131500244},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.09480488300323486}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/percomworkshops53856.2022.9767442","is_oa":false,"landing_page_url":"https://doi.org/10.1109/percomworkshops53856.2022.9767442","pdf_url":null,"source":{"id":"https://openalex.org/S4363608020","display_name":"2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops)","raw_type":"proceedings-article"},{"id":"pmh:oai:eprints.soton.ac.uk:491463","is_oa":false,"landing_page_url":"https://eprints.soton.ac.uk/491463/","pdf_url":null,"source":{"id":"https://openalex.org/S4306401019","display_name":"ePrints Soton (University of Southampton)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I43439940","host_organization_name":"University of Southampton","host_organization_lineage":["https://openalex.org/I43439940"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Conference or Workshop Item"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":28,"referenced_works":["https://openalex.org/W2112796928","https://openalex.org/W2119144962","https://openalex.org/W2659864996","https://openalex.org/W2798170643","https://openalex.org/W2883780447","https://openalex.org/W2950865323","https://openalex.org/W2953384591","https://openalex.org/W2963122961","https://openalex.org/W2963163009","https://openalex.org/W3004066795","https://openalex.org/W3038006402","https://openalex.org/W3104263540","https://openalex.org/W3116942696","https://openalex.org/W3166185557","https://openalex.org/W3177113534","https://openalex.org/W3196677734","https://openalex.org/W3210540241","https://openalex.org/W4287124550","https://openalex.org/W4288365403","https://openalex.org/W4295312788","https://openalex.org/W6677580257","https://openalex.org/W6713134421","https://openalex.org/W6746698991","https://openalex.org/W6753767121","https://openalex.org/W6762407294","https://openalex.org/W6766978945","https://openalex.org/W6774015895","https://openalex.org/W6800984526"],"related_works":["https://openalex.org/W1657880117","https://openalex.org/W2595172197","https://openalex.org/W2127970246","https://openalex.org/W2887888236","https://openalex.org/W2146894470","https://openalex.org/W1555620806","https://openalex.org/W2038675678","https://openalex.org/W4229055905","https://openalex.org/W4225434596","https://openalex.org/W2185570015"],"abstract_inverted_index":{"Deep":[0,25],"Learning":[1],"(DL)":[2],"has":[3,50],"shown":[4],"impressive":[5],"performance":[6,177],"in":[7,174],"many":[8],"mobile":[9,32,74,81],"applications.":[10],"Most":[11],"existing":[12],"works":[13],"have":[14],"focused":[15],"on":[16,30,47,71,93,100,106,143,161,164,198],"reducing":[17],"the":[18,35,66,79,125,130,144,153,175,193,199],"computational":[19],"and":[20,44,85,108,119,201,209],"resource":[21],"overheads":[22],"of":[23,38,68,178,188],"running":[24,90],"Neural":[26],"Networks":[27],"(DNN)":[28],"inference":[29],"resource-constrained":[31],"devices.":[33],"However,":[34,149],"other":[36],"aspect":[37],"DNN":[39,155],"operations,":[40],"i.e.":[41],"training":[42,70,99,156,189],"(forward":[43],"backward":[45],"passes)":[46],"smartphone":[48],"GPUs,":[49],"received":[51],"little":[52],"attention":[53],"thus":[54],"far.":[55],"To":[56,123],"this":[57,115],"end,":[58],"we":[59,96,128,150,206],"conduct":[60],"an":[61],"initial":[62],"analysis":[63],"to":[64,114,192],"examine":[65],"feasibility":[67],"on-device":[69],"smartphones":[72],"using":[73],"GPUs.":[75,94],"We":[76],"first":[77],"employ":[78],"open-source":[80],"DL":[82],"framework":[83],"(MNN)":[84],"its":[86],"OpenCL":[87,131],"backend":[88],"for":[89,212],"compute":[91],"kernels":[92],"Next,":[95],"observed":[97],"that":[98,152,167],"CPUs":[101,140],"is":[102,157],"much":[103,159],"faster":[104],"than":[105,163],"GPUs":[107,162],"identified":[109],"two":[110],"possible":[111],"bottlenecks":[112],"related":[113],"observation:":[116],"(i)":[117],"computation":[118,126],"(ii)":[120],"memory":[121,168],"bottlenecks.":[122],"solve":[124],"bottleneck,":[127],"optimize":[129],"backend’s":[132],"kernels,":[133],"showing":[134],"2x":[135],"improvements":[136],"(40-70":[137],"GFLOPs)":[138,142],"over":[139,180],"(15-30":[141],"Snapdragon":[145],"8":[146],"series":[147],"processors.":[148],"find":[151],"full":[154],"still":[158],"slower":[160],"CPUs,":[165],"indicating":[166],"bottleneck":[169],"plays":[170],"a":[171],"significant":[172],"role":[173],"lower":[176],"GPU":[179],"CPU.":[181],"The":[182],"data":[183],"movement":[184],"takes":[185],"almost":[186],"91%":[187],"time":[190],"due":[191],"low":[194],"bandwidth.":[195],"Lastly,":[196],"based":[197],"findings":[200],"failures":[202],"during":[203],"our":[204],"investigation,":[205],"present":[207],"limitations":[208],"practical":[210],"guidelines":[211],"future":[213],"directions.":[214]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":7},{"year":2023,"cited_by_count":3},{"year":2022,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}