{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T11:42:48Z","timestamp":1784288568454,"version":"3.55.0"},"reference-count":81,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"9","license":[{"start":{"date-parts":[[2023,9,1]],"date-time":"2023-09-01T00:00:00Z","timestamp":1693526400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2023,9,1]],"date-time":"2023-09-01T00:00:00Z","timestamp":1693526400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2023,9,1]],"date-time":"2023-09-01T00:00:00Z","timestamp":1693526400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001866","name":"National Research Fund (FNR), Luxembourg","doi-asserted-by":"publisher","award":["R-AGR-0424-05-D\/Bjorn Ottersten"],"award-info":[{"award-number":["R-AGR-0424-05-D\/Bjorn Ottersten"]}],"id":[{"id":"10.13039\/501100001866","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001866","name":"National Research Fund (FNR), Luxembourg","doi-asserted-by":"publisher","award":["CPPP17\/IS\/11643091\/IDform\/Aouada"],"award-info":[{"award-number":["CPPP17\/IS\/11643091\/IDform\/Aouada"]}],"id":[{"id":"10.13039\/501100001866","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001866","name":"National Research Fund (FNR), Luxembourg","doi-asserted-by":"publisher","award":["BRIDGES2020\/IS\/14755859\/MEET-A\/Aouada"],"award-info":[{"award-number":["BRIDGES2020\/IS\/14755859\/MEET-A\/Aouada"]}],"id":[{"id":"10.13039\/501100001866","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Neural Netw. Learning Syst."],"published-print":{"date-parts":[[2023,9]]},"DOI":"10.1109\/tnnls.2021.3131813","type":"journal-article","created":{"date-parts":[[2022,1,5]],"date-time":"2022-01-05T20:32:11Z","timestamp":1641414731000},"page":"5961-5975","source":"Crossref","is-referenced-by-count":43,"title":["Why Is Everyone Training Very Deep Neural Network With Skip Connections?"],"prefix":"10.1109","volume":"34","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4652-1691","authenticated-orcid":false,"given":"Oyebade K.","family":"Oyedotun","sequence":"first","affiliation":[{"name":"Interdisciplinary Centre for Security, Reliability, Trust (SnT), University of Luxembourg, Luxembourg"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kassem Al","family":"Ismaeil","sequence":"additional","affiliation":[{"name":"Interdisciplinary Centre for Security, Reliability, Trust (SnT), University of Luxembourg, Luxembourg"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7576-2064","authenticated-orcid":false,"given":"Djamila","family":"Aouada","sequence":"additional","affiliation":[{"name":"Interdisciplinary Centre for Security, Reliability, Trust (SnT), University of Luxembourg, Luxembourg"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1016\/j.compeleceng.2018.01.036"},{"key":"ref57","first-page":"2902","article-title":"Deep linear networks with arbitrary loss: All local minima are global","author":"laurent","year":"2018","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-29911-8_20"},{"key":"ref56","first-page":"3727","article-title":"Optimization landscape and expressivity of deep CNNs","author":"nguyen","year":"2018","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ICSAI48974.2019.9010178"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1007\/BF02288367"},{"key":"ref14","first-page":"1","article-title":"Nearly-tight VC-dimension and pseudodimension bounds for piecewise linear neural networks","volume":"20","author":"bartlett","year":"2019","journal-title":"J Mach Learn Res"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1134\/S1064562410040010"},{"key":"ref53","first-page":"31","article-title":"Transport analysis of infinitely deep neural network","volume":"20","author":"sonoda","year":"2019","journal-title":"J Mach Learn Res"},{"key":"ref52","first-page":"1","article-title":"Critical points of linear neural networks: Analytical forms and landscape properties","author":"zhou","year":"2019","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2017.09.007"},{"key":"ref55","first-page":"586","article-title":"Deep learning without poor local minima","author":"kawaguchi","year":"2016","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref10","first-page":"1","article-title":"Visualizing deep neural network decisions: Prediction difference analysis","author":"zintgraf","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref54","first-page":"1","article-title":"Exact solutions to the nonlinear dynamics of learning in deep linear neural networks","author":"saxe","year":"2014","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref17","first-page":"6048","article-title":"Evolutionary stochastic gradient descent for optimization of deep neural networks","author":"cui","year":"2018","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref16","first-page":"2847","article-title":"On the expressive power of deep neural networks","author":"raghu","year":"2017","journal-title":"Proc 34th Int Conf Mach Learn"},{"key":"ref19","first-page":"6155","article-title":"Learning and generalization in overparameterized neural networks, going beyond two layers","author":"allen-zhu","year":"2019","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2017.06.003"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v32i1.11668"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1088\/1361-6420\/aa9a90"},{"key":"ref46","first-page":"1","article-title":"Highway and residual networks learn unrolled iterative estimation","author":"greff","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref45","first-page":"550","article-title":"Residual networks behave like ensembles of relatively shallow networks","author":"veit","year":"2016","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v32i1.11680"},{"key":"ref47","first-page":"1","article-title":"Residual connections encourage iterative inference","author":"jastrzebski","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2017.2654543"},{"key":"ref41","first-page":"448","article-title":"Batch normalization: Accelerating deep network training by reducing internal covariate shift","author":"ioffe","year":"2015","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref44","first-page":"342","article-title":"The shattered gradients problem: If ResNets are the answer, then what is the question?","author":"balduzzi","year":"2017","journal-title":"Proc 34th Int Conf Mach Learn"},{"key":"ref43","first-page":"4467","article-title":"Dual path networks","author":"chen","year":"2017","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref49","first-page":"1","article-title":"Multi-level residual networks from dynamical systems view","author":"chang","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref8","first-page":"1","article-title":"Learning how to explain neural networks: Patternnet and patternattribution","author":"kindermans","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref7","first-page":"1675","article-title":"Gradient descent finds global minima of deep neural networks","author":"du","year":"2019","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2016.2599820"},{"key":"ref4","first-page":"1","article-title":"iNNvestigate neural networks","volume":"20","author":"alber","year":"2019","journal-title":"J Mach Learn Res"},{"key":"ref3","first-page":"1","article-title":"Towards reverse-engineering black-box neural networks","author":"oh","year":"2018","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref6","first-page":"1","article-title":"Gradient descent provably optimizes over-parameterized neural networks","author":"du","year":"2019","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v33i01.33013681"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1093\/biomet\/38.3-4.481"},{"key":"ref40","first-page":"249","article-title":"Understanding the difficulty of training deep feedforward neural networks","author":"glorot","year":"2010","journal-title":"Proc 13th Int Conf Artif Intell Statist"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1017\/S0266466600004746"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.415"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.1093\/biomet\/60.2.415"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.634"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1080\/00029890.1979.11994929"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v31i1.11231"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.243"},{"key":"ref31","first-page":"1","article-title":"Very deep convolutional networks for large-scale image recognition","author":"simonyan","year":"2015","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref75","first-page":"1","article-title":"Darts: Differentiable architecture search","author":"liu","year":"2019","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref30","first-page":"2377","article-title":"Training very deep networks","author":"srivastava","year":"2015","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref74","article-title":"Network in network","author":"lin","year":"2013","journal-title":"arXiv 1312 4400"},{"key":"ref33","first-page":"1","article-title":"FractalNet: Ultra-deep neural networks without residuals","author":"larsson","year":"2017","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref77","first-page":"1","article-title":"Understanding and robustifying differentiable architecture search","author":"zela","year":"2020","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref32","first-page":"630","article-title":"Identity mappings in deep residual networks","author":"he","year":"2016","journal-title":"Proc Eur Conf Comput Vis"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00138"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/JBHI.2017.2775662"},{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2017.05.025"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.123"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-70096-0_3"},{"key":"ref71","article-title":"Highway networks","author":"srivastava","year":"2015","journal-title":"arXiv 1505 00387"},{"key":"ref70","first-page":"5393","article-title":"Dynamical isometry and a mean field theory of CNNs: How to train 10,000-layer vanilla convolutional neural networks","author":"xiao","year":"2018","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref73","first-page":"1319","article-title":"Maxout networks","author":"goodfellow","year":"2013","journal-title":"Proc Int Conf Mach Learn"},{"key":"ref72","first-page":"1","article-title":"Striving for simplicity: The all convolutional net","author":"springenberg","year":"2015","journal-title":"Proc Int Conf Learn Represent Workshop"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2010.936030"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2012.02.023"},{"key":"ref67","first-page":"35","article-title":"Wide residual networks","author":"zagoruyko","year":"2016","journal-title":"Proc Brit Mach Vis Conf"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2013.2293637"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW.2018.00217"},{"key":"ref69","article-title":"DiracNets: Training very deep neural networks without skip-connections","author":"zagoruyko","year":"2017","journal-title":"arXiv 1706 00388"},{"key":"ref20","first-page":"10835","article-title":"Generalization bounds of stochastic gradient descent for wide and deep neural networks","author":"cao","year":"2019","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref64","first-page":"1576","article-title":"Non-asymptotic theory of random matrices: Extreme singular values","author":"rudelson","year":"2010","journal-title":"Proc Int Congr Mathematicians"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2004.12.011"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1007\/s00521-016-2294-8"},{"key":"ref66","author":"krizhevsky","year":"2019","journal-title":"Cifar-10 and cifar- 100 (canadian institute for advanced research)"},{"key":"ref21","first-page":"1","article-title":"Predicting the generalization gap in deep networks with margin distributions","author":"jiang","year":"2019","journal-title":"Proc Int Conf Learn Represent"},{"key":"ref65","author":"lecun","year":"2019","journal-title":"MNIST Handwritten Digit Database"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2019.01.006"},{"key":"ref27","first-page":"666","article-title":"Shallow vs. deep sum-product networks","author":"delalleau","year":"2011","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.1016\/0885-064X(87)90027-6"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1007\/BF01932285"},{"key":"ref61","first-page":"202","article-title":"Avoiding pathologies in very deep networks","author":"duvenaud","year":"2014","journal-title":"Artificial Intelligence and Statistics"}],"container-title":["IEEE Transactions on Neural Networks and Learning Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/5962385\/10237282\/09669990.pdf?arnumber=9669990","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,9,25]],"date-time":"2023-09-25T18:24:00Z","timestamp":1695666240000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9669990\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,9]]},"references-count":81,"journal-issue":{"issue":"9"},"URL":"https:\/\/doi.org\/10.1109\/tnnls.2021.3131813","relation":{},"ISSN":["2162-237X","2162-2388"],"issn-type":[{"value":"2162-237X","type":"print"},{"value":"2162-2388","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,9]]}}}