{"id":"https://openalex.org/W4408354156","doi":"https://doi.org/10.1109/icassp49660.2025.10887669","title":"Concentrating Harder for Faster Audio Transformer","display_name":"Concentrating Harder for Faster Audio Transformer","publication_year":2025,"publication_date":"2025-03-12","ids":{"openalex":"https://openalex.org/W4408354156","doi":"https://doi.org/10.1109/icassp49660.2025.10887669"},"language":"en","primary_location":{"id":"doi:10.1109/icassp49660.2025.10887669","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49660.2025.10887669","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","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/A5080714748","display_name":"Lars E. Schmidt","orcid":null},"institutions":[{"id":"https://openalex.org/I4210123192","display_name":"International Audio Laboratories Erlangen","ror":"https://ror.org/02mkz3e80","country_code":"DE","type":"facility","lineage":["https://openalex.org/I181369854","https://openalex.org/I4210123192","https://openalex.org/I4210124274","https://openalex.org/I4923324"]}],"countries":["DE"],"is_corresponding":true,"raw_author_name":"Lorenz P. Schmidt","raw_affiliation_strings":["Friedrich-Alexander-Universitat Erlangen-Nürnberg,International Audio Laboratories,Germany"],"affiliations":[{"raw_affiliation_string":"Friedrich-Alexander-Universitat Erlangen-Nürnberg,International Audio Laboratories,Germany","institution_ids":["https://openalex.org/I4210123192"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101869282","display_name":"Nils Peters","orcid":"https://orcid.org/0000-0001-6758-4491"},"institutions":[{"id":"https://openalex.org/I205274468","display_name":"Trinity College Dublin","ror":"https://ror.org/02tyrky19","country_code":"IE","type":"education","lineage":["https://openalex.org/I205274468"]}],"countries":["IE"],"is_corresponding":false,"raw_author_name":"Nils Peters","raw_affiliation_strings":["Trinity College,Dublin,Ireland"],"affiliations":[{"raw_affiliation_string":"Trinity College,Dublin,Ireland","institution_ids":["https://openalex.org/I205274468"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5080714748"],"corresponding_institution_ids":["https://openalex.org/I4210123192"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.04758605,"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":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12564","display_name":"Sensor Technology and Measurement Systems","score":0.9394999742507935,"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"}},"topics":[{"id":"https://openalex.org/T12564","display_name":"Sensor Technology and Measurement Systems","score":0.9394999742507935,"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/computer-science","display_name":"Computer science","score":0.690135657787323},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.47164982557296753},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.2764630913734436},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.13832440972328186},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.057389408349990845}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.690135657787323},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.47164982557296753},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.2764630913734436},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.13832440972328186},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.057389408349990845}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp49660.2025.10887669","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49660.2025.10887669","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.8199999928474426,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":15,"referenced_works":["https://openalex.org/W2593116425","https://openalex.org/W3014596384","https://openalex.org/W3092557781","https://openalex.org/W3113822960","https://openalex.org/W3196974791","https://openalex.org/W4206281850","https://openalex.org/W4306955484","https://openalex.org/W4307977385","https://openalex.org/W4390678101","https://openalex.org/W6748865747","https://openalex.org/W6762673008","https://openalex.org/W6774324527","https://openalex.org/W6843734122","https://openalex.org/W6848208918","https://openalex.org/W6854308872"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Attention-based":[0],"models":[1],"have":[2],"become":[3],"tremendously":[4],"successful":[5],"in":[6,38,46,195],"the":[7,78,127,133,142,153,172,177],"last":[8],"couple":[9],"of":[10,29,85,109],"years":[11],"for":[12,89,103,116,164,171],"tasks":[13],"such":[14,125],"as":[15,126],"Acoustic":[16,136],"Scene":[17],"Classification,":[18,20],"Event":[19],"or":[21],"Speaker":[22],"Identification.":[23],"They":[24],"operate":[25],"on":[26,82],"a":[27,55,59,74,83,192],"set":[28],"tokens":[30,111],"extracted":[31],"from":[32],"audio":[33,94,123],"features":[34],"and":[35,66,92,112,120,141,174],"scale":[36],"quadratically":[37],"sequence":[39],"length":[40],"due":[41],"to":[42,53,62,80],"their":[43],"pairwise":[44],"operations":[45],"multi-head":[47],"attention.In":[48],"this":[49],"paper,":[50],"we":[51,189],"propose":[52],"parameterize":[54],"Dirichlet":[56],"prior":[57],"with":[58,184,200],"Transformer":[60],"model,":[61],"jointly":[63],"estimate":[64],"label":[65],"token":[67,70,104,114,154,178,186],"probabilities.":[68],"A":[69],"bottleneck,":[71],"provided":[72],"by":[73],"Categorical-Dirichlet":[75],"pair,":[76],"forces":[77],"model":[79],"concentrate":[81],"subset":[84],"tokens.":[86],"This":[87],"allows":[88],"improved":[90],"interpretability":[91],"higher":[93],"throughput":[95,196],"during":[96],"inference.":[97],"We":[98,168],"compare":[99],"two":[100],"different":[101],"methods":[102],"sampling":[105,115],"\u2013":[106],"full":[107],"knowledge":[108],"all":[110],"partial":[113,185],"reduced":[117,158],"complexity.We":[118],"evaluate":[119],"interpret":[121],"typical":[122],"datasets":[124],"Environmental":[128],"Sound":[129],"Classification":[130],"(ESC-50)":[131],"dataset,":[132,140],"TAU":[134],"Urban":[135],"Scenes":[137],"2020":[138],"(TAU20)":[139],"Speech":[143],"Command":[144],"version":[145],"2":[146],"(SCv2)":[147],"dataset.":[148],"The":[149],"results":[150],"show":[151,169,190],"that":[152,170,191],"budget":[155],"can":[156,180,197],"be":[157,181,198],"without":[159],"significant":[160,193],"performance":[161],"loss,":[162],"especially":[163],"acoustic":[165],"scene":[166],"classification.":[167],"ESC-50":[173],"SCv2":[175],"datasets,":[176],"relevance":[179],"well":[182],"approximated":[183],"view.":[187],"Finally,":[188],"increase":[194],"achieved":[199],"our":[201],"proposed":[202],"methods.":[203]},"counts_by_year":[],"updated_date":"2025-12-22T23:10:17.713674","created_date":"2025-10-10T00:00:00"}