{"id":"https://openalex.org/W4410583102","doi":"https://doi.org/10.23919/date64628.2025.10992745","title":"DyLGNN: Efficient LM-GNN Fine-Tuning with Dynamic Node Partitioning, Low-Degree Sparsity, and Asynchronous Sub-Batch","display_name":"DyLGNN: Efficient LM-GNN Fine-Tuning with Dynamic Node Partitioning, Low-Degree Sparsity, and Asynchronous Sub-Batch","publication_year":2025,"publication_date":"2025-03-31","ids":{"openalex":"https://openalex.org/W4410583102","doi":"https://doi.org/10.23919/date64628.2025.10992745"},"language":"en","primary_location":{"id":"doi:10.23919/date64628.2025.10992745","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10992745","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 Design, Automation & Test in Europe Conference (DATE)","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/A5100523306","display_name":"Zhen Yu","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Zhen Yu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068894182","display_name":"Jinhao Li","orcid":"https://orcid.org/0009-0009-4286-6359"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jinhao Li","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111878216","display_name":"Jianrong Xu","orcid":"https://orcid.org/0009-0000-7000-6537"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiaming Xu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109113870","display_name":"Shan Huang","orcid":"https://orcid.org/0009-0005-2012-8540"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shan Huang","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066342192","display_name":"J. Ye","orcid":"https://orcid.org/0009-0007-6189-3251"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiancai Ye","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018513988","display_name":"Ningyi Xu","orcid":"https://orcid.org/0009-0004-6809-7694"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ningyi Xu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5015946486","display_name":"Guohao Dai","orcid":"https://orcid.org/0000-0003-0849-3252"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guohao Dai","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5100523306"],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.11537926,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9941999912261963,"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"}},"topics":[{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9941999912261963,"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/T10502","display_name":"Advanced Memory and Neural Computing","score":0.989799976348877,"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/T10472","display_name":"Semiconductor materials and devices","score":0.913100004196167,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/asynchronous-communication","display_name":"Asynchronous communication","score":0.8242971897125244},{"id":"https://openalex.org/keywords/degree","display_name":"Degree (music)","score":0.6752405166625977},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6748886108398438},{"id":"https://openalex.org/keywords/node","display_name":"Node (physics)","score":0.6010855436325073},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.14657780528068542},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.06020447611808777}],"concepts":[{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.8242971897125244},{"id":"https://openalex.org/C2775997480","wikidata":"https://www.wikidata.org/wiki/Q586277","display_name":"Degree (music)","level":2,"score":0.6752405166625977},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6748886108398438},{"id":"https://openalex.org/C62611344","wikidata":"https://www.wikidata.org/wiki/Q1062658","display_name":"Node (physics)","level":2,"score":0.6010855436325073},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.14657780528068542},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.06020447611808777},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C24890656","wikidata":"https://www.wikidata.org/wiki/Q82811","display_name":"Acoustics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.23919/date64628.2025.10992745","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10992745","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 Design, Automation & Test in Europe Conference (DATE)","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":19,"referenced_works":["https://openalex.org/W1980867644","https://openalex.org/W2914721378","https://openalex.org/W2949911880","https://openalex.org/W2997128522","https://openalex.org/W2997261254","https://openalex.org/W3007364240","https://openalex.org/W3034202788","https://openalex.org/W4385568277","https://openalex.org/W4392367398","https://openalex.org/W4396735838","https://openalex.org/W4401024755","https://openalex.org/W4402683078","https://openalex.org/W4403792254","https://openalex.org/W4407953340","https://openalex.org/W4409282403","https://openalex.org/W6699364125","https://openalex.org/W6738964360","https://openalex.org/W6776488958","https://openalex.org/W6838322825"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2116677773","https://openalex.org/W2155261584","https://openalex.org/W2399035783","https://openalex.org/W2584231425","https://openalex.org/W2150611273","https://openalex.org/W2390279801","https://openalex.org/W4391913857"],"abstract_inverted_index":{"Text-Attributed":[0],"Graphs":[1],"(TAGs)":[2],"tasks":[3],"involve":[4],"both":[5],"textual":[6],"node":[7,267],"information":[8],"and":[9,23,38,88,170,280,294,309,325],"graph":[10,29,307],"topological":[11],"structure.":[12],"The":[13],"top-k":[14,188,252,333],"method,":[15,189],"using":[16],"Language":[17],"Models":[18],"(LMs)":[19],"for":[20,28,78,181,287],"text":[21,122],"encoding":[22],"Graph":[24],"Neural":[25],"Networks":[26],"(GNNs)":[27],"processing,":[30],"offers":[31],"the":[32,92,109,121,129,132,143,178,186,192,206,215,234,251,256,260,284,312,332],"best":[33],"accuracy":[34,71,220],"while":[35],"balancing":[36],"memory":[37,137,194,272,323],"training":[39,179,193],"time.":[40,111,144],"However,":[41],"challenges":[42],"still":[43],"exist:":[44],"(1)":[45,160],"Static":[46],"sampling":[47,60],"of":[48,94,108,142,208,259],"k":[49,56],"neighbors":[50],"reduces":[51],"performance.":[52],"Using":[53],"a":[54,69,319,326],"fixed":[55],"can":[57,275],"result":[58],"in":[59,105,177,296,322],"too":[61,64],"few":[62],"or":[63],"many":[65],"nodes,":[66,103,119,239],"leading":[67],"to":[68,138,175,185,231,331],"3.2%":[70],"loss":[72],"across":[73],"datasets.":[74],"(2)":[75,198],"Time-consuming":[76,113],"processing":[77,117],"non-trainable":[79,90,95,118],"nodes.":[80],"After":[81],"partitioning":[82],"all":[83],"nodes":[84,96,174,210,230],"into":[85,126],"with-gradient":[86],"trainable":[87,102],"without-gradient":[89],"sets,":[91],"number":[93],"is":[97],"~9-10":[98],"x":[99,244],"larger":[100],"than":[101],"resulting":[104],"nearly":[106],"70%":[107],"total":[110],"(3)":[112,246],"data":[114,133,278,288],"movement.":[115,289],"For":[116],"after":[120],"strings":[123],"are":[124],"tokenized":[125],"tokens":[127],"on":[128,214,228,304],"CPU":[130],"side,":[131],"movement":[134,279],"from":[135],"host":[136],"GPU":[139,292],"takes":[140],"30%-40%":[141],"In":[145],"this":[146],"paper,":[147],"we":[148,190,254,274],"propose":[149,165],"DyLGNN,":[150],"an":[151,166,241,297],"efficient":[152],"end-to-end":[153,328],"LM-GNN":[154],"fine-tuning":[155],"framework":[156],"through":[157],"three":[158,313],"innovations:":[159],"Heuristic":[161],"Node":[162],"Partitioning.":[163],"We":[164,202,223,301],"algorithm":[167],"that":[168,205],"dynamically":[169],"adaptively":[171],"selects":[172],"\u201cimportant\u201d":[173],"participate":[176],"process":[180],"downstream":[182],"tasks.":[183],"Compared":[184],"static":[187],"reduce":[191,233],"usage":[195,324],"by":[196,237,310],"24.0%.":[197],"Low-Degree":[199],"Sparse":[200],"Attention.":[201],"point":[203],"out":[204],"embedding":[207],"low-degree":[209,229],"has":[211],"minimal":[212],"impact":[213],"final":[216],"results":[217,295],"(e.g.":[218],"~1.5%":[219],"loss),":[221],"therefore,":[222],"perform":[224],"sparse":[225],"attention":[226],"computation":[227,235],"further":[232],"caused":[236],"\u201cunimportant\u201d":[238],"achieving":[240],"average":[242,298],"1.27":[243],"speedup.":[245,300],"Asynchronous":[247],"Sub-batch":[248],"Pipeline.":[249],"Within":[250],"framework,":[253],"analyze":[255],"time":[257,285],"breakdown":[258],"LM":[261],"inference":[262],"component.":[263],"Leveraging":[264],"our":[265],"heuristic":[266],"partitioning,":[268],"which":[269],"effectively":[270],"minimizes":[271],"demands,":[273],"asynchronously":[276],"execute":[277],"computation,":[281],"thereby":[282],"overlapping":[283],"required":[286],"This":[290],"improves":[291],"utilization":[293],"1.1x":[299],"conduct":[302],"experiments":[303],"several":[305],"common":[306],"datasets,":[308],"combining":[311],"methods":[314],"mentioned":[315],"above,":[316],"DyLGNN":[317],"achieves":[318],"22.0%":[320],"reduction":[321],"1.3x":[327],"speedup":[329],"compared":[330],"strategy.":[334]},"counts_by_year":[],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}