{"id":"https://openalex.org/W7150970794","doi":"https://doi.org/10.48550/arxiv.2604.02651","title":"Communication-free Sampling and 4D Hybrid Parallelism for Scalable Mini-batch GNN Training","display_name":"Communication-free Sampling and 4D Hybrid Parallelism for Scalable Mini-batch GNN Training","publication_year":2026,"publication_date":"2026-04-03","ids":{"openalex":"https://openalex.org/W7150970794","doi":"https://doi.org/10.48550/arxiv.2604.02651"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.02651","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.02651","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.02651","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5039652101","display_name":"Cunyang Wei","orcid":"https://orcid.org/0009-0001-8910-4951"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Cunyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133033718","display_name":"Siddharth Singh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Singh, Siddharth","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060450571","display_name":"Aishwarya Sarkar","orcid":"https://orcid.org/0000-0003-2575-1241"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sarkar, Aishwarya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068348488","display_name":"Daniel Nichols","orcid":"https://orcid.org/0000-0002-3538-6164"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nichols, Daniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133041799","display_name":"Tisha Patel","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Patel, Tisha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133034059","display_name":"Aditya K. Ranjan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ranjan, Aditya K.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133014207","display_name":"Sayan Ghosh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ghosh, Sayan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133013878","display_name":"Ali Jannesari","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jannesari, Ali","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047746050","display_name":"Nathan R. Tallent","orcid":"https://orcid.org/0000-0003-4297-3057"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tallent, Nathan R.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5081506338","display_name":"Abhinav Bhatel\u00e9","orcid":"https://orcid.org/0000-0003-3069-3701"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bhatele, Abhinav","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":10,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.8499000072479248,"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"}},"topics":[{"id":"https://openalex.org/T11273","display_name":"Advanced Graph Neural Networks","score":0.8499000072479248,"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/T12292","display_name":"Graph Theory and Algorithms","score":0.0851999968290329,"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/T14347","display_name":"Big Data and Digital Economy","score":0.007899999618530273,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/scalability","display_name":"Scalability","score":0.7092000246047974},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6270999908447266},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5192000269889832},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.49900001287460327},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.49720001220703125},{"id":"https://openalex.org/keywords/data-parallelism","display_name":"Data parallelism","score":0.4853000044822693},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.4837000072002411},{"id":"https://openalex.org/keywords/kernel","display_name":"Kernel (algebra)","score":0.43849998712539673}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8547999858856201},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.7092000246047974},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6270999908447266},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.576200008392334},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5192000269889832},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.49900001287460327},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.49720001220703125},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.4853000044822693},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.4837000072002411},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.43849998712539673},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.36629998683929443},{"id":"https://openalex.org/C17349429","wikidata":"https://www.wikidata.org/wiki/Q1049914","display_name":"Matrix multiplication","level":3,"score":0.35910001397132874},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.3560999929904938},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.3483999967575073},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.3467999994754791},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.34610000252723694},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34549999237060547},{"id":"https://openalex.org/C80899671","wikidata":"https://www.wikidata.org/wiki/Q1304193","display_name":"Vertex (graph theory)","level":3,"score":0.3240000009536743},{"id":"https://openalex.org/C2779599972","wikidata":"https://www.wikidata.org/wiki/Q82231","display_name":"NoSQL","level":3,"score":0.32179999351501465},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.28929999470710754},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2669999897480011},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.2662000060081482},{"id":"https://openalex.org/C106515295","wikidata":"https://www.wikidata.org/wiki/Q26806595","display_name":"Parallel processing","level":2,"score":0.2606000006198883},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.25690001249313354},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2517000138759613}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.02651","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.02651","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.02651","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.02651","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Graph":[0],"neural":[1],"networks":[2],"(GNNs)":[3],"are":[4],"widely":[5],"used":[6],"for":[7,33,67],"learning":[8],"on":[9,156,168,172,177,191],"graph":[10,158],"datasets":[11,159],"derived":[12],"from":[13,18],"various":[14],"real-world":[15],"scenarios.":[16],"Learning":[17],"extremely":[19],"large":[20],"graphs":[21],"requires":[22],"distributed":[23,38,75],"training,":[24,138],"and":[25,50,82,150,160,174],"mini-batching":[26],"with":[27,124,137],"sampling":[28,48,90,136],"is":[29],"a":[30,63,87],"popular":[31],"approach":[32],"parallelizing":[34],"GNN":[35,70],"training.":[36],"Existing":[37],"mini-batch":[39,69,113],"approaches":[40],"have":[41],"significant":[42],"performance":[43],"bottlenecks":[44],"due":[45],"to":[46,97,115,134,165],"expensive":[47],"methods":[49],"limited":[51],"scaling":[52,112,163],"when":[53],"using":[54],"data":[55,83,122,144],"parallelism.":[56,84],"In":[57],"this":[58],"work,":[59],"we":[60],"present":[61,131],"ScaleGNN,":[62],"4D":[64],"parallel":[65,78],"framework":[66],"scalable":[68],"training":[71,114,185],"that":[72],"combines":[73],"communication-free":[74],"sampling,":[76],"3D":[77,109],"matrix":[79],"multiplication":[80],"(PMM),":[81],"ScaleGNN":[85,155,181],"introduces":[86],"uniform":[88],"vertex":[89],"algorithm,":[91],"enabling":[92],"each":[93],"process":[94],"(GPU":[95],"device)":[96],"construct":[98],"its":[99],"local":[100],"mini-batch,":[101],"i.e.,":[102],"subgraph":[103],"partitions":[104],"without":[105],"any":[106],"inter-process":[107],"communication.":[108],"PMM":[110],"enables":[111],"much":[116],"larger":[117],"GPU":[118],"counts":[119],"than":[120],"vanilla":[121],"parallelism":[123],"significantly":[125],"lower":[126,146],"communication":[127,140],"overheads.":[128],"We":[129,153],"also":[130],"additional":[132],"optimizations":[133],"overlap":[135],"reduce":[139],"overhead":[141],"by":[142],"sending":[143],"in":[145],"precision,":[147],"kernel":[148],"fusion,":[149],"communication-computation":[151],"overlap.":[152],"evaluate":[154],"five":[157],"demonstrate":[161],"strong":[162],"up":[164],"2048":[166,170],"GPUs":[167,176],"Perlmutter,":[169,180],"GCDs":[171],"Frontier,":[173],"1024":[175],"Tuolumne.":[178],"On":[179],"achieves":[182],"3.5x":[183],"end-to-end":[184],"speedup":[186],"over":[187],"the":[188],"SOTA":[189],"baseline":[190],"ogbn-products.":[192]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-07T00:00:00"}