{"id":"https://openalex.org/W4414359653","doi":"https://doi.org/10.24963/ijcai.2025/730","title":"Learning Advanced Self-Attention for Linear Transformers in the Singular Value Domain","display_name":"Learning Advanced Self-Attention for Linear Transformers in the Singular Value Domain","publication_year":2025,"publication_date":"2025-09-01","ids":{"openalex":"https://openalex.org/W4414359653","doi":"https://doi.org/10.24963/ijcai.2025/730"},"language":"en","primary_location":{"id":"doi:10.24963/ijcai.2025/730","is_oa":false,"landing_page_url":"https://doi.org/10.24963/ijcai.2025/730","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","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/A5067239392","display_name":"Hyowon Wi","orcid":"https://orcid.org/0009-0000-7954-3128"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":true,"raw_author_name":"Hyowon Wi","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST)","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5057709875","display_name":"Jeongwhan Choi","orcid":"https://orcid.org/0000-0002-6530-2662"},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jeongwhan Choi","raw_affiliation_strings":["Yonsei University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Yonsei University","institution_ids":["https://openalex.org/I193775966"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5067253588","display_name":"Noseong Park","orcid":"https://orcid.org/0000-0002-1268-840X"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Noseong Park","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST)","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5067239392"],"corresponding_institution_ids":["https://openalex.org/I157485424"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.12195094,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"6561","last_page":"6569"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.6306999921798706,"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/T10320","display_name":"Neural Networks and Applications","score":0.6306999921798706,"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/adjacency-matrix","display_name":"Adjacency matrix","score":0.6775000095367432},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.44110000133514404},{"id":"https://openalex.org/keywords/filter","display_name":"Filter (signal processing)","score":0.3855000138282776},{"id":"https://openalex.org/keywords/signal-processing","display_name":"Signal processing","score":0.3749000132083893},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.34310001134872437},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.3345000147819519},{"id":"https://openalex.org/keywords/signal-flow-graph","display_name":"Signal-flow graph","score":0.3125}],"concepts":[{"id":"https://openalex.org/C180356752","wikidata":"https://www.wikidata.org/wiki/Q727035","display_name":"Adjacency matrix","level":3,"score":0.6775000095367432},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.49869999289512634},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4921000003814697},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.44110000133514404},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4122999906539917},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.40310001373291016},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.3855000138282776},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.3749000132083893},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.34310001134872437},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3345000147819519},{"id":"https://openalex.org/C166501922","wikidata":"https://www.wikidata.org/wiki/Q1786523","display_name":"Signal-flow graph","level":2,"score":0.3125},{"id":"https://openalex.org/C22597639","wikidata":"https://www.wikidata.org/wiki/Q5449227","display_name":"Filter design","level":3,"score":0.30979999899864197},{"id":"https://openalex.org/C146380142","wikidata":"https://www.wikidata.org/wiki/Q1137726","display_name":"Directed graph","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C19118579","wikidata":"https://www.wikidata.org/wiki/Q786423","display_name":"Frequency domain","level":2,"score":0.2799000144004822},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.2687000036239624},{"id":"https://openalex.org/C110484373","wikidata":"https://www.wikidata.org/wiki/Q264398","display_name":"Adjacency list","level":2,"score":0.26840001344680786},{"id":"https://openalex.org/C139722471","wikidata":"https://www.wikidata.org/wiki/Q2619517","display_name":"Linear filter","level":3,"score":0.26030001044273376},{"id":"https://openalex.org/C134727501","wikidata":"https://www.wikidata.org/wiki/Q5597073","display_name":"Graph bandwidth","level":5,"score":0.25270000100135803}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.24963/ijcai.2025/730","is_oa":false,"landing_page_url":"https://doi.org/10.24963/ijcai.2025/730","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Transformers":[0,12],"have":[1,29],"demonstrated":[2],"remarkable":[3],"performance":[4,167],"across":[5],"diverse":[6],"domains.":[7],"The":[8],"key":[9],"component":[10],"of":[11,42,49,87,101,143],"is":[13,77,181],"self-attention,":[14],"which":[15],"learns":[16],"the":[17,24,32,47,54,68,72,75,84,88,98,128,132,136,141,151,155],"relationship":[18],"between":[19],"any":[20],"two":[21],"tokens":[22],"in":[23,111,135],"input":[25,156],"sequence.":[26],"Recent":[27],"studies":[28],"revealed":[30],"that":[31,163],"self-attention":[33,55,76,107,129],"can":[34,56],"be":[35,57],"understood":[36],"as":[37,60,71,93,130],"a":[38,43,61,78,94,112,119],"normalized":[39],"adjacency":[40],"matrix":[41],"graph.":[44],"Notably,":[45],"from":[46,140],"perspective":[48,142],"graph":[50,63,79,133,144],"signal":[51,145],"processing":[52,146],"(GSP),":[53],"equivalently":[58],"defined":[59,81],"simple":[62],"filter,":[64],"applying":[65],"GSP":[66],"using":[67],"value":[69,138],"vector":[70],"signal.":[73],"However,":[74],"filter":[80,96,134],"with":[82,150],"only":[83],"first":[85],"order":[86],"polynomial":[89],"matrix,":[90],"and":[91,176],"acts":[92],"low-pass":[95],"preventing":[97],"effective":[99],"leverage":[100],"various":[102,169],"frequency":[103],"information.":[104],"Consequently,":[105],"existing":[106],"mechanisms":[108],"are":[109],"designed":[110],"rather":[113],"simplified":[114],"manner.":[115],"Therefore,":[116],"we":[117,161],"propose":[118],"novel":[120],"method,":[121],"called":[122],"Attentive":[123],"Graph":[124],"Filter":[125],"(AGF),":[126],"interpreting":[127],"learning":[131],"singular":[137],"domain":[139],"for":[147],"directed":[148],"graphs":[149],"linear":[152],"complexity":[153],"w.r.t.":[154],"length.":[157],"In":[158],"our":[159],"experiments,":[160],"demonstrate":[162],"AGF":[164],"achieves":[165],"state-of-the-art":[166],"on":[168],"tasks,":[170],"including":[171],"Long":[172],"Range":[173],"Arena":[174],"benchmark":[175],"time":[177],"series":[178],"classification.":[179],"Code":[180],"available":[182],"at":[183],"https://github.com/hyowonwi/agf.":[184]},"counts_by_year":[],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}