{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T03:08:57Z","timestamp":1772593737076,"version":"3.50.1"},"reference-count":91,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"5","license":[{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"am","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/100007863","name":"K2I Graduate Fellowship (BP) from Rice University","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100007863","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000001","name":"NSF","doi-asserted-by":"publisher","award":["CCF-1911094"],"award-info":[{"award-number":["CCF-1911094"]}],"id":[{"id":"10.13039\/100000001","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000001","name":"NSF","doi-asserted-by":"publisher","award":["IIS-1838177"],"award-info":[{"award-number":["IIS-1838177"]}],"id":[{"id":"10.13039\/100000001","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000001","name":"NSF","doi-asserted-by":"publisher","award":["IIS-1730574"],"award-info":[{"award-number":["IIS-1730574"]}],"id":[{"id":"10.13039\/100000001","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000183","name":"Army Research Office","doi-asserted-by":"publisher","award":["W911NF-15-1-0316"],"award-info":[{"award-number":["W911NF-15-1-0316"]}],"id":[{"id":"10.13039\/100000183","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000181","name":"Air Force Office of Scientific Research","doi-asserted-by":"publisher","award":["FA9550-14-1-0088"],"award-info":[{"award-number":["FA9550-14-1-0088"]}],"id":[{"id":"10.13039\/100000181","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000181","name":"Air Force Office of Scientific Research","doi-asserted-by":"publisher","award":["FA9550-18-1-0478"],"award-info":[{"award-number":["FA9550-18-1-0478"]}],"id":[{"id":"10.13039\/100000181","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000006","name":"Office of Naval Research","doi-asserted-by":"publisher","award":["N00014-17-1-2551"],"award-info":[{"award-number":["N00014-17-1-2551"]}],"id":[{"id":"10.13039\/100000006","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000006","name":"Office of Naval Research","doi-asserted-by":"publisher","award":["N00014-18-12571"],"award-info":[{"award-number":["N00014-18-12571"]}],"id":[{"id":"10.13039\/100000006","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000185","name":"Defense Advanced Research Projects Agency","doi-asserted-by":"publisher","award":["G001534-7500"],"award-info":[{"award-number":["G001534-7500"]}],"id":[{"id":"10.13039\/100000185","id-type":"DOI","asserted-by":"publisher"}]},{"name":"DOD Vannevar Bush Faculty Fellowship of ONR","award":["N00014-18-1-2047"],"award-info":[{"award-number":["N00014-18-1-2047"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Proc. IEEE"],"published-print":{"date-parts":[[2021,5]]},"DOI":"10.1109\/jproc.2020.3042100","type":"journal-article","created":{"date-parts":[[2020,12,17]],"date-time":"2020-12-17T20:34:34Z","timestamp":1608237274000},"page":"704-727","source":"Crossref","is-referenced-by-count":28,"title":["Mad Max: Affine Spline Insights Into Deep Learning"],"prefix":"10.1109","volume":"109","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-5692-4187","authenticated-orcid":false,"given":"Randall","family":"Balestriero","sequence":"first","affiliation":[{"name":"Department of Electrical and Computer Engineering, Rice University, Houston, TX, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0721-8999","authenticated-orcid":false,"given":"Richard G.","family":"Baraniuk","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, Rice University, Houston, TX, USA"}]}],"member":"263","reference":[{"key":"ref73","first-page":"15832","article-title":"The geometry of deep networks: Power diagram subdivision","author":"balestriero","year":"2019","journal-title":"Proc Neural Inf Process Syst (NIPS)"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1016\/j.patrec.2009.09.011"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.21236\/ADA307097"},{"key":"ref70","first-page":"3","article-title":"Autoencoders, minimum description length and Helmholtz free energy","author":"hinton","year":"1994","journal-title":"Proc Adv Neural Inf Process Syst (NIPS)"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1117\/12.351328"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1117\/12.366780"},{"key":"ref74","article-title":"Max-affine spline insights into deep generative networks","author":"balestriero","year":"2020","journal-title":"arXiv 2002 11912"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1145\/1390156.1390294"},{"key":"ref75","article-title":"Implicit rugosity regularization via data augmentation","author":"lejeune","year":"2019","journal-title":"arXiv 1905 11639"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/18.256506"},{"key":"ref78","first-page":"1050","article-title":"Dropout as a Bayesian approximation: Representing model uncertainty in deep learning","author":"gal","year":"2016","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref79","first-page":"261","article-title":"Learning algorithms for classification: A comparison on handwritten digit recognition","author":"lecun","year":"1995","journal-title":"Neural Networks"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1016\/B978-0-444-89597-4.50013-5"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.2514\/3.8935"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4612-6333-3"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9781139171502"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/TIT.2005.859246"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1016\/S0895-7177(99)00195-8"},{"key":"ref35","article-title":"Accurate piecewise linear continuous approximations to one-dimensional curves: Error estimates and algorithms","author":"nishikawa","year":"1998"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9780511804441"},{"key":"ref60","article-title":"Adam: A method for stochastic optimization","author":"kingma","year":"2014","journal-title":"arXiv 1412 6980"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1109\/26.3776"},{"key":"ref61","volume":"159","author":"gersho","year":"2012","journal-title":"Vector Quantization and Signal Compression"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1145\/344779.345009"},{"key":"ref28","doi-asserted-by":"crossref","DOI":"10.1093\/oso\/9780198538493.001.0001","author":"bishop","year":"1995","journal-title":"Neural Networks for Pattern Recognition"},{"key":"ref64","article-title":"Compressing deep convolutional networks using vector quantization","author":"gong","year":"2014","journal-title":"arXiv 1412 6115"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.123"},{"key":"ref65","article-title":"Soft-to-hard vector quantization for end-to-end learning compressible representations","author":"agustsson","year":"2017","journal-title":"arXiv 1704 00648"},{"key":"ref66","article-title":"Deep learning vector quantization","author":"de vries","year":"2016","journal-title":"Proc ESANN"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1016\/B978-0-12-741252-8.50010-8"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1016\/j.aeue.2016.12.002"},{"key":"ref68","first-page":"2579","article-title":"DeepVQ: A deep network architecture for vector quantization","author":"le tan","year":"2018","journal-title":"Proc CVPR Workshops"},{"key":"ref2","first-page":"698","article-title":"On the expressive power of deep learning: A tensor analysis","volume":"49","author":"cohen","year":"2016","journal-title":"Proc 29th Annu Conf Learn Theory"},{"key":"ref69","article-title":"A minimum description length framework for unsupervised learning","author":"zemel","year":"1994"},{"key":"ref1","first-page":"818","article-title":"Visualizing and understanding convolutional networks","author":"zeiler","year":"2014","journal-title":"Proc Eur Conf Comput Vis (ECCV)"},{"key":"ref20","first-page":"2924","article-title":"On the number of linear regions of deep neural networks","author":"montufar","year":"2014","journal-title":"Proc Adv Neural Inf Process Syst (NIPS)"},{"key":"ref22","article-title":"A representer theorem for deep neural networks","author":"unser","year":"2018","journal-title":"arXiv 1802 09210"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2017.06.009"},{"key":"ref24","first-page":"374","article-title":"A spline theory of deep networks","author":"balestriero","year":"2018","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref23","article-title":"Understanding deep learning requires rethinking generalization","author":"zhang","year":"2016","journal-title":"arXiv 1611 03530"},{"key":"ref26","article-title":"Mad max: Affine spline insights into deep learning","author":"balestriero","year":"2018","journal-title":"arXiv 1805 06576"},{"key":"ref25","volume":"1","author":"goodfellow","year":"2016","journal-title":"Deep Learning"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.3115\/v1\/D14-1179"},{"key":"ref51","article-title":"Empirical evaluation of gated recurrent neural networks on sequence modeling","author":"chung","year":"2014","journal-title":"arXiv 1412 3555"},{"key":"ref91","article-title":"Nonlinear approximation and (Deep) ReLU networks","author":"daubechies","year":"2019","journal-title":"arXiv 1905 02199"},{"key":"ref90","article-title":"Universal approximation power of deep neural networks via nonlinear control theory","author":"tabuada","year":"2020","journal-title":"arXiv 2007 06007"},{"key":"ref59","first-page":"424","article-title":"Local decorrelation for improved pedestrian detection","author":"nam","year":"2014","journal-title":"Proc Adv Neural Inf Process Syst (NIPS)"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2001.940237"},{"key":"ref57","article-title":"Adversarial examples in the physical world","author":"kurakin","year":"2016","journal-title":"arXiv 1607 02533"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1997.9.1.1"},{"key":"ref55","article-title":"Batch normalization: Accelerating deep network training by reducing internal covariate shift","author":"ioffe","year":"2015","journal-title":"arXiv 1502 03167"},{"key":"ref54","article-title":"Deep inside convolutional networks: Visualising image classification models and saliency maps","author":"simonyan","year":"2013","journal-title":"arXiv 1312 6034"},{"key":"ref53","article-title":"Intriguing properties of neural networks","author":"szegedy","year":"2013","journal-title":"arXiv 1312 6199"},{"key":"ref52","author":"rabiner","year":"1975","journal-title":"Theory and Application of Digital Signal Processing"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/ITW.2015.7133169"},{"key":"ref11","first-page":"1947","article-title":"Emergence of invariance and disentanglement in deep representations","volume":"19","author":"achille","year":"2018","journal-title":"J Mach Learn Res"},{"key":"ref40","first-page":"2672","article-title":"Generative adversarial nets","author":"goodfellow","year":"2014","journal-title":"Proc Int Conf Neural Inf Process Syst (NIPS)"},{"key":"ref12","author":"lecun","year":"1998","journal-title":"The MNIST Database of Handwritten Digits"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref14","first-page":"2377","article-title":"Training very deep networks","author":"srivastava","year":"2015","journal-title":"Proc Adv Neural Inf Process Syst (NIPS)"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/72.159058"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2013.6639349"},{"key":"ref16","article-title":"Generating sequences with recurrent neural networks","author":"graves","year":"2013","journal-title":"arXiv 1308 0850 [cs]"},{"key":"ref81","doi-asserted-by":"crossref","first-page":"533","DOI":"10.1038\/323533a0","article-title":"Learning representations by back-propagating errors","volume":"323","author":"rumelhart","year":"1986","journal-title":"Nature"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v31i1.11231"},{"key":"ref84","first-page":"2121","article-title":"Adaptive subgradient methods for online learning and stochastic optimization","volume":"12","author":"duchi","year":"2011","journal-title":"J Mach Learn Res"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1007\/s11081-008-9045-3"},{"key":"ref83","article-title":"ADADELTA: An adaptive learning rate method","author":"zeiler","year":"2012","journal-title":"arXiv 1212 5701"},{"key":"ref19","first-page":"3261","article-title":"Multivariate convex regression with adaptive partitioning","volume":"14","author":"hannah","year":"2013","journal-title":"J Mach Learn Res"},{"key":"ref80","article-title":"An exploration of softmax alternatives belonging to the spherical loss family","author":"de br\u00e9bisson","year":"2015","journal-title":"arXiv 1511 05042"},{"key":"ref4","article-title":"Exponentially vanishing sub-optimal local minima in multilayer neural networks","author":"soudry","year":"2017","journal-title":"arXiv 1702 05777"},{"key":"ref89","first-page":"6169","article-title":"ResNet with one-neuron hidden layers is a universal approximator","author":"lin","year":"2018","journal-title":"Proc Adv Neural Inf Process Syst"},{"key":"ref3","article-title":"Depth creates no bad local minima","author":"lu","year":"2017","journal-title":"arXiv 1702 08580"},{"key":"ref6","first-page":"1","article-title":"Convolutional neural networks analyzed via convolutional sparse coding","volume":"18","author":"papyan","year":"2017","journal-title":"J Mach Learn Res"},{"key":"ref5","article-title":"Visual representations: Defining properties and deep approximations","author":"soatto","year":"2016","journal-title":"Proc Int Conf Learn Represent (ICLR)"},{"key":"ref8","article-title":"Provable bounds for learning some deep representations","author":"arora","year":"2013","journal-title":"arXiv 1310 6343"},{"key":"ref85","first-page":"26","article-title":"Lecture 6.5-RMSprop: Divide the gradient by a running average of its recent magnitude","volume":"4","author":"tieleman","year":"2012","journal-title":"COURSERA Neural Netw Mach Learn"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2012.230"},{"key":"ref86","author":"mallat","year":"1999","journal-title":"A Wavelet Tour of Signal Processing"},{"key":"ref49","article-title":"Maxout networks","author":"goodfellow","year":"2013","journal-title":"arXiv 1302 4389"},{"key":"ref87","author":"jayaraman","year":"2009","journal-title":"Digital Image Processing"},{"key":"ref9","first-page":"2558","article-title":"A probabilistic framework for deep learning","author":"patel","year":"2016","journal-title":"Proc Neural Inf Process Syst (NIPS)"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN.2005.1556215"},{"key":"ref46","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 (NIPS)"},{"key":"ref45","article-title":"Resnet in resnet: Generalizing residual architectures","author":"targ","year":"2016","journal-title":"arXiv 1603 08029"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1007\/BF02551274"},{"key":"ref47","article-title":"A max-affine spline perspective of recurrent neural networks","author":"wang","year":"2019","journal-title":"Proc Int Conf Learn Represent (ICLR)"},{"key":"ref42","first-page":"146","article-title":"Input convex neural networks","author":"amos","year":"2017","journal-title":"Proc Int Conf Mach Learn (ICML)"},{"key":"ref41","article-title":"Understanding deep neural networks with rectified linear units","author":"arora","year":"2016","journal-title":"arXiv 1611 01491"},{"key":"ref44","article-title":"From hard to soft: Understanding deep network nonlinearities via vector quantization and statistical inference","author":"balestriero","year":"2019","journal-title":"Proc Int Conf Learn Represent (ICLR)"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1137\/120887795"}],"container-title":["Proceedings of the IEEE"],"original-title":[],"link":[{"URL":"https:\/\/ieeexplore.ieee.org\/ielam\/5\/9420072\/9296823-aam.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/5\/9420072\/09296823.pdf?arnumber=9296823","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,12,11]],"date-time":"2024-12-11T01:33:09Z","timestamp":1733880789000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9296823\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,5]]},"references-count":91,"journal-issue":{"issue":"5"},"URL":"https:\/\/doi.org\/10.1109\/jproc.2020.3042100","relation":{},"ISSN":["0018-9219","1558-2256"],"issn-type":[{"value":"0018-9219","type":"print"},{"value":"1558-2256","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021,5]]}}}