{"id":"https://openalex.org/W7002186285","doi":null,"title":"Moment Dynamics in Self-Normalising Neural Networks","display_name":"Moment Dynamics in Self-Normalising Neural Networks","publication_year":2017,"publication_date":"2017-01-01","ids":{"openalex":"https://openalex.org/W7002186285"},"language":"en","primary_location":{"id":"pmh:oai:epub.jku.at/:2344947","doi":null,"is_oa":true,"landing_page_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","pdf_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","source":{"id":"https://openalex.org/S4306401788","display_name":"University Library Linz repository (Johannes Kepler Universitat Linz)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I121883995","host_organization_name":"Johannes Kepler University of Linz","host_organization_lineage":["https://openalex.org/I121883995"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Thesis"},"volume":null,"type":"dissertation","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Hoedt, Pieter-Jan","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Hoedt, Pieter-Jan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":true,"primary_topic":{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.9717000126838684,"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":0.9717000126838684,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.016300000250339508,"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":0.002099999925121665,"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/backward-pass","display_name":"backward pass","score":0.6769999861717224},{"id":"https://openalex.org/keywords/data-whitening","display_name":"data whitening","score":0.7350000143051147},{"id":"https://openalex.org/keywords/weight-initialization","display_name":"weight initialization","score":0.8309999704360962},{"id":"https://openalex.org/keywords/learning-rate-schedules","display_name":"learning rate scheduling","score":0.8320000171661377},{"id":"https://openalex.org/keywords/internal-covariate-shift","display_name":"internal covariate shift","score":0.7559999823570251},{"id":"https://openalex.org/keywords/neural-network-architecture","display_name":"neural network architecture","score":0.5320000052452087},{"id":"https://openalex.org/keywords/deep-neural-network","display_name":"deep neural network","score":0.43700000643730164},{"id":"https://openalex.org/keywords/vanishing-gradient","display_name":"gradient vanishing","score":0.6309999823570251},{"id":"https://openalex.org/keywords/hierarchical-representation","display_name":"hierarchical representation","score":0.38499999046325684},{"id":"https://openalex.org/keywords/representation-learning","display_name":"representation learning","score":0.4390000104904175}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6468999981880188},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5928000211715698},{"id":"https://openalex.org/C179254644","wikidata":"https://www.wikidata.org/wiki/Q13222844","display_name":"Moment (physics)","level":2,"score":0.5860000252723694},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.4771000146865845},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41269999742507935},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35580000281333923},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.34049999713897705},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.2768000066280365},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.26649999618530273},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2635999917984009}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:epub.jku.at/:2344947","doi":null,"is_oa":true,"landing_page_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","pdf_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","source":{"id":"https://openalex.org/S4306401788","display_name":"University Library Linz repository (Johannes Kepler Universitat Linz)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I121883995","host_organization_name":"Johannes Kepler University of Linz","host_organization_lineage":["https://openalex.org/I121883995"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Thesis"}],"best_oa_location":{"id":"pmh:oai:epub.jku.at/:2344947","doi":null,"is_oa":true,"landing_page_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","pdf_url":"http://resolver.obvsg.at/urn:nbn:at:at-ubl:1-19148","source":{"id":"https://openalex.org/S4306401788","display_name":"University Library Linz repository (Johannes Kepler Universitat Linz)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I121883995","host_organization_name":"Johannes Kepler University of Linz","host_organization_lineage":["https://openalex.org/I121883995"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Thesis"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7002186285.pdf","grobid_xml":"https://content.openalex.org/works/W7002186285.grobid-xml"},"referenced_works_count":22,"referenced_works":["https://openalex.org/W6908809","https://openalex.org/W1994197834","https://openalex.org/W1995341919","https://openalex.org/W2001771035","https://openalex.org/W2038507439","https://openalex.org/W2064675550","https://openalex.org/W2069143585","https://openalex.org/W2076063813","https://openalex.org/W2112796928","https://openalex.org/W2123045220","https://openalex.org/W2136848157","https://openalex.org/W2136922672","https://openalex.org/W2144499799","https://openalex.org/W2194775991","https://openalex.org/W2384495648","https://openalex.org/W2525778437","https://openalex.org/W2624413595","https://openalex.org/W2919115771","https://openalex.org/W2964121744","https://openalex.org/W3037932933","https://openalex.org/W3198350258","https://openalex.org/W4238404964"],"related_works":["https://openalex.org/W2624413595","https://openalex.org/W3133475568","https://openalex.org/W4398768831","https://openalex.org/W2997791832","https://openalex.org/W3022080084","https://openalex.org/W3000400623","https://openalex.org/W3035766514","https://openalex.org/W3167550568","https://openalex.org/W2982280733","https://openalex.org/W2795358072"],"abstract_inverted_index":{"In":[0],"the":[1,23,67,112,122,132,138,141,164,171,181,185,190,201,220,224,254,257,271,275,281,310,313,322,337,349,353,372,410,442,483,486,509,543,550,562,566],"passed":[2],"decade,":[3],"deep":[4,18,26,60,96,437],"learning":[5,13,19,59,94,227,247,323,329,438,460],"has":[6,73,317,393,462,476,513,564],"achieved":[7],"state-of-the-art":[8],"performance":[9,225],"for":[10,219,240,298,339,470,504,508,517,549,583,591],"various":[11],"machine":[12,246],"tasks.The":[14],"wide":[15],"applicability":[16],"of":[17,25,51,146,173,183,229,256,312,342,366,371,444,478,485,538,554,561],"mainly":[20],"arises":[21],"from":[22],"ability":[24],"neural":[27,40,61,152,448],"networks":[28,153,208,266,449],"to":[29,83,106,120,189,207,233,347,359,368,380,389,408,414,455,466,481,494,519,534,558,579],"learn":[30],"useful":[31,465,548],"features":[32],"by":[33,76,110,159],"themselves.By":[34],"combining":[35],"multiple":[36,586],"layers":[37,123,294],"in":[38,58,95,114,124,140,187,193,265,292,450,568],"a":[39,125,364,378],"network,":[41],"hierarchical":[42],"representations":[43],"can":[44,156,295],"be":[45,157,296],"created":[46],"with":[47,209,231,261,267,289,307,377],"an":[48,286],"increasing":[49],"level":[50],"abstraction.However,":[52],"there":[53],"is":[54,64,119],"one":[55],"fundamental":[56],"difficulty":[57],"networks,":[62],"which":[63,143,194,453],"known":[65],"as":[66,100,137,218,488,540],"vanishing":[68],"gradient":[69],"problem.Today,":[70],"this":[71,108,177,398,471,555,592],"issue":[72,109],"been":[74,104,463,477,546],"alleviated":[75],"new":[77,402],"activation":[78],"functions":[79],"and":[80,135,155,198,212,226,242,283,325,384,400,431,446,501,542,571,588],"better":[81],"ways":[82],"initialise":[84],"weights.Although":[85],"less":[86],"severe,":[87],"also":[88,374,406,492],"internal":[89],"covariate":[90],"shift":[91],"slows":[92],"down":[93],"networks.Several":[97],"techniques":[98],"such":[99],"batch":[101],"normalisation":[102],"have":[103,131,545],"proposed":[105],"counter":[107],"normalising":[111],"data":[113,139,235,249],"each":[115,167],"layer.An":[116],"alternative":[117],"approach":[118],"construct":[121],"network":[126,282],"so":[127,148],"that":[128,166,214,252,285,303,326,355,412],"their":[129,530],"activations":[130],"same":[133],"mean":[134],"variance":[136,255,291],"input.Networks":[142],"are":[144,149],"capable":[145],"doing":[147],"called":[150],"Self-Normalising":[151],"(SNNs)":[154],"constructed":[158],"enforcing":[160],"certain":[161],"characteristics":[162],"onto":[163],"mappings":[165],"layer":[168],"induces":[169],"on":[170,244,321,521,529],"moments":[172,186],"its":[174],"input":[175],"data.In":[176],"thesis,":[178],"we":[179,195,204,222],"extend":[180,205],"idea":[182],"analysing":[184],"SNNs":[188,206,230,304],"backward":[191,202],"pass,":[192],"both":[196],"theoretically":[197],"empirically":[199],"investigate":[200],"dynamics.Further,":[203],"bias":[210],"units":[211],"show":[213],"similar":[215],"dynamics":[216],"hold":[217],"weights.Additionally,":[221],"compare":[223],"behaviour":[228,324],"respect":[232],"different":[234,458],"normalization":[236],"techniques,":[237],"weight":[238,315],"distributions":[239],"initialization":[241],"optimisers":[243],"several":[245],"benchmarking":[248],"sets.We":[250],"find":[251],"(1)":[253],"weights":[258],"steadily":[259],"increases":[260],"very":[262,464],"small":[263],"steps":[264],"random":[268],"errors,":[269],"(2)":[270],"architecture":[272],"affects":[273],"how":[274],"error":[276,287],"signal":[277,288],"propagates":[278],"back":[279],"through":[280,397],"(3)":[284],"reduced":[290],"lower":[293],"advantageous":[297],"learning.Furthermore,":[299],"our":[300],"analysis":[301],"reveals":[302],"perform":[305],"best":[306],"whitened":[308],"data,":[309],"choice":[311],"initial":[314],"distribution":[316],"no":[318],"significant":[319],"effect":[320],"most":[327,560],"adaptive":[328],"rate":[330],"schedules":[331],"do":[332],"help,":[333],"although":[334],"they":[335,489],"break":[336],"conditions":[338],"self-normalisation.i":[340],"First":[341],"all,":[343],"I":[344,532,576],"would":[345,405,491,577],"like":[346,407,418,493,578],"thank":[348,409,496,580],"bioinformatics":[350],"institute":[351],"at":[352],"JKU":[354],"provided":[356,375,401],"me":[357,362,376,396,518],"access":[358],"GPUs.This":[360],"saved":[361],"quite":[363],"bit":[365],"time":[367],"run":[369],"all":[370],"experiments.They":[373],"forum":[379],"present":[381],"my":[382,569],"ideas":[383,403],"retrieve":[385],"valuable":[386,589],"feedback.Special":[387],"thanks":[388],"G\u00fcnter":[390],"Klambauer":[391],"who":[392],"patiently":[394],"guided":[395],"thesis":[399],".I":[404],"people":[411],"contribute(d)":[413],"open":[415],"source":[416],"projects":[417],"Tensorflow":[419],"(Google)":[420],"(Abadi":[421],"et":[422,428],"al.,":[423,429],"2015),":[424],"Theano":[425],"(MILA)":[426],"(Al-Rfou":[427],"2016)":[430],"Keras":[432],"(Fran\u00e7ois":[433],"Chollet)":[434],"(Chollet,":[435],"2015).These":[436],"frameworks":[439],"greatly":[440],"simplified":[441],"process":[443],"building":[445],"training":[447],"python.Especially":[451],"Keras,":[452],"allows":[454],"switch":[456],"between":[457],"lower-level":[459],"frameworks,":[461],"set":[467],"up":[468],"experiments":[469,487],"thesis.Also":[472],"matplotlib":[473],"(Hunter,":[474],"2007)":[475],"great":[479],"use":[480],"make":[482],"visualisations":[484],"are.I":[490],"explicitly":[495],"St\u00e9fan":[497],"van":[498],"der":[499],"Walt":[500],"Nathaniel":[502],"Smith":[503],"developing":[505],"virridis.Their":[506],"quest":[507],"perfect":[510],"colour":[511,525],"map":[512],"made":[514],"it":[515],"easier":[516],"focus":[520],"content":[522],"rather":[523],"than":[524],"schemes.For":[526],"more":[527,551],"information":[528],"work,":[531],"refer":[533],"https://bids.github.io/colormap/.The":[535],"(free)":[536],"availability":[537],"platforms":[539],"wolframalpha":[541],"wolframcloud":[544],"extremely":[547],"theoretical":[552],"parts":[553],"thesis.Being":[556],"able":[557],"double-check":[559],"equations":[563],"solidified":[565],"confidence":[567],"research":[570],"writing.Last,":[572],"but":[573],"not":[574],"least,":[575],"Katharina":[581],"Prinz":[582],"her":[584],"patience,":[585],"reviews":[587],"feedback":[590],"thesis.v":[593],"5.1":[594],"MNIST":[595],"images":[596],".":[597,598]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-10-10T00:00:00"}