{"id":"https://openalex.org/W7161173263","doi":"https://doi.org/10.48550/arxiv.2605.13612","title":"Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning","display_name":"Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning","publication_year":2026,"publication_date":"2026-05-13","ids":{"openalex":"https://openalex.org/W7161173263","doi":"https://doi.org/10.48550/arxiv.2605.13612"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2605.13612","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.13612","pdf_url":"https://arxiv.org/pdf/2605.13612","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2605.13612","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5039205039","display_name":"Yatin Dandi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Dandi, Yatin","raw_affiliation_strings":["Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","Statistical Physics of Computation Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]},{"raw_affiliation_string":"Statistical Physics of Computation Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092061419","display_name":"Matteo Vilucchio","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Vilucchio, Matteo","raw_affiliation_strings":["Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101959142","display_name":"Luca Arnaboldi","orcid":"https://orcid.org/0009-0001-9739-8849","observed_orcids":["https://orcid.org/0009-0001-9739-8849"]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Arnaboldi, Luca","raw_affiliation_strings":["Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125760334","display_name":"Hugo Tabanelli","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Tabanelli, Hugo","raw_affiliation_strings":["Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5068236230","display_name":"Florent Krz\u0105ka\u0142a","orcid":"https://orcid.org/0000-0003-2313-2578","observed_orcids":["https://orcid.org/0000-0003-2313-2578"]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Krzakala, Florent","raw_affiliation_strings":["Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Learning and Physics Laboratory , \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL)","institution_ids":["https://openalex.org/I5124864"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I5124864"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":2,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.5259000062942505,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.5259000062942505,"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/T10320","display_name":"Neural Networks and Applications","score":0.42800000309944153,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.00839999970048666,"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/hierarchical-representation-learning","display_name":"hierarchical feature learning","score":0.8629999756813049},{"id":"https://openalex.org/keywords/lazy-training","display_name":"lazy training","score":0.3840000033378601},{"id":"https://openalex.org/keywords/concept-emergence","display_name":"concept emergence","score":0.5350000262260437},{"id":"https://openalex.org/keywords/sample-complexity","display_name":"sample complexity","score":0.5220000147819519},{"id":"https://openalex.org/keywords/compositional-features","display_name":"compositional features","score":0.4560000002384186},{"id":"https://openalex.org/keywords/convolutional-neural-networks","display_name":"convolutional neural networks","score":0.7120000123977661},{"id":"https://openalex.org/keywords/random-feature-models","display_name":"random feature models","score":0.34200000762939453}],"concepts":[{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.765999972820282},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6984999775886536},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.6952000260353088},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5879999995231628},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.5353000164031982},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5181000232696533},{"id":"https://openalex.org/C38935604","wikidata":"https://www.wikidata.org/wiki/Q4330363","display_name":"Stylized fact","level":2,"score":0.4661000072956085},{"id":"https://openalex.org/C112313634","wikidata":"https://www.wikidata.org/wiki/Q7886648","display_name":"Complement (music)","level":5,"score":0.43479999899864197},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.4255000054836273},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.41190001368522644},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.38999998569488525},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.385699987411499},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.37369999289512634},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.35089999437332153},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.35040000081062317},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.337799996137619},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3028999865055084},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.2833000123500824},{"id":"https://openalex.org/C111030470","wikidata":"https://www.wikidata.org/wiki/Q1430460","display_name":"Curse of dimensionality","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C60908668","wikidata":"https://www.wikidata.org/wiki/Q690207","display_name":"Perceptron","level":3,"score":0.2605000138282776}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2605.13612","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.13612","pdf_url":"https://arxiv.org/pdf/2605.13612","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2605.13612","doi":"https://doi.org/10.48550/arxiv.2605.13612","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13612","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2605.13612","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.13612","pdf_url":"https://arxiv.org/pdf/2605.13612","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2365339206","display_name":"OPtimal Estimation in RAndom Generative mOdelS:  Theory, Thresholds, Tractability","funder_award_id":"200390","funder_id":"https://openalex.org/F4320320924","funder_display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung"},{"id":"https://openalex.org/G874360974","display_name":"DSGIANGO: The DynamicS of GradIent Descent AlGOrithm in Neural Networks","funder_award_id":"225837","funder_id":"https://openalex.org/F4320320924","funder_display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320320924","display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung","ror":"https://ror.org/00yjd3n13"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7161173263.pdf","grobid_xml":"https://content.openalex.org/works/W7161173263.grobid-xml"},"referenced_works_count":16,"referenced_works":["https://openalex.org/W3157286395","https://openalex.org/W4385473649","https://openalex.org/W4396914181","https://openalex.org/W4404306248","https://openalex.org/W4407759146","https://openalex.org/W4412420048","https://openalex.org/W4415134327","https://openalex.org/W4416150179","https://openalex.org/W4417028086","https://openalex.org/W7083807327","https://openalex.org/W7106337870","https://openalex.org/W7125927084","https://openalex.org/W7127446538","https://openalex.org/W7128073161","https://openalex.org/W7128504444","https://openalex.org/W7128679019"],"related_works":["https://openalex.org/W7135071062","https://openalex.org/W4388481916","https://openalex.org/W3000127803","https://openalex.org/W4313303785","https://openalex.org/W4383859664","https://openalex.org/W7133227840","https://openalex.org/W3098527778","https://openalex.org/W4415797820","https://openalex.org/W4415882150","https://openalex.org/W7160651178"],"abstract_inverted_index":{"Understanding":[0],"how":[1,105,113],"deep":[2,20,79],"neural":[3],"networks":[4],"learn":[5],"useful":[6],"internal":[7],"representations":[8,106,168],"from":[9,134],"data":[10],"remains":[11],"a":[12,29,74,83,90,124],"central":[13],"open":[14],"problem":[15],"in":[16,35],"the":[17,49,56,59,70,100,142],"theory":[18,143],"of":[19,32,115],"learning.":[21],"We":[22,140],"introduce":[23],"Neural":[24,87,155],"Low-Degree":[25],"Filtering":[26],"(Neural":[27],"LoFi),":[28],"stylized":[30],"limit":[31],"gradient-based":[33],"training":[34],"which":[36,128],"hierarchical":[37],"feature":[38,97,173],"learning":[39,98],"becomes":[40],"an":[41],"explicit":[42,92],"iterative":[43],"spectral":[44],"procedure.":[45],"In":[46],"this":[47],"limit,":[48],"dynamics":[50],"at":[51],"each":[52],"layer":[53,61,109],"decouple:":[54],"given":[55,119],"current":[57],"representation,":[58],"next":[60],"selects":[62],"directions":[63],"with":[64,82,118,144,170,175],"maximal":[65],"accessible":[66],"low-degree":[67,138],"correlation":[68],"to":[69],"label.":[71],"This":[72],"yields":[73],"tractable":[75],"surrogate":[76],"mechanism":[77,126],"for":[78,94],"learning,":[80],"together":[81],"natural":[84],"kernel-space":[85],"interpretation.":[86],"LoFi":[88,156],"provides":[89],"mathematically":[91],"framework":[93],"studying":[95],"multi-layer":[96],"beyond":[99],"lazy":[101,159],"regime.":[102],"It":[103],"predicts":[104,167],"are":[107],"selected":[108],"by":[110,127],"layer,":[111],"explains":[112],"emergence":[114],"concepts":[116],"arises":[117],"sample":[120],"complexity,":[121],"and":[122,150,166],"gives":[123],"concrete":[125],"depth":[129],"progressively":[130],"constructs":[131],"new":[132],"features":[133],"old":[135],"ones":[136],"through":[137],"compositionality.":[139],"complement":[141],"mechanistic":[145],"experiments":[146],"on":[147],"fully":[148],"connected":[149],"convolutional":[151],"architectures,":[152],"showing":[153],"that":[154],"improves":[157],"over":[158],"random-feature":[160],"baselines,":[161],"recovers":[162],"meaningful":[163],"structured":[164],"filters,":[165],"aligned":[169],"early":[171],"gradient-descent":[172],"discovery":[174],"real":[176],"datasets.":[177]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-05-15T00:00:00"}