{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T19:25:59Z","timestamp":1775503559857,"version":"3.50.1"},"reference-count":24,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/tdm\/userlicense\/1.0\/"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/legal\/tdmrep-license"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-017"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-012"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-004"}],"content-domain":{"domain":["elsevier.com","sciencedirect.com"],"crossmark-restriction":true},"short-container-title":["Neural Networks"],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.1016\/j.neunet.2026.108559","type":"journal-article","created":{"date-parts":[[2026,1,8]],"date-time":"2026-01-08T16:52:43Z","timestamp":1767891163000},"page":"108559","update-policy":"https:\/\/doi.org\/10.1016\/elsevier_cm_policy","source":"Crossref","is-referenced-by-count":1,"special_numbering":"C","title":["Two-hidden-layer ReLU neural networks and finite elements"],"prefix":"10.1016","volume":"198","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-2169-1491","authenticated-orcid":false,"given":"Pengzhan","family":"Jin","sequence":"first","affiliation":[]}],"member":"78","reference":[{"key":"10.1016\/j.neunet.2026.108559_bib0001","unstructured":"Arora, R., Basu, A., Mianjy, P., & Mukherjee, A. (2016). Understanding deep neural networks with rectified linear units. ICLR 2018. arXiv preprint arXiv: 1611.01491."},{"key":"10.1016\/j.neunet.2026.108559_bib0002","series-title":"The mathematical theory of finite element methods","author":"Brenner","year":"2008"},{"key":"10.1016\/j.neunet.2026.108559_bib0003","doi-asserted-by":"crossref","first-page":"7167","DOI":"10.52202\/068431-0520","article-title":"Improved bounds on neural complexity for representing piecewise linear functions","volume":"35","author":"Chen","year":"2022","journal-title":"Advances in Neural Information Processing Systems"},{"key":"10.1016\/j.neunet.2026.108559_bib0004","series-title":"The finite element method for elliptic problems","author":"Ciarlet","year":"2002"},{"issue":"4","key":"10.1016\/j.neunet.2026.108559_bib0005","doi-asserted-by":"crossref","first-page":"303","DOI":"10.1007\/BF02551274","article-title":"Approximation by superpositions of a sigmoidal function","volume":"2","author":"Cybenko","year":"1989","journal-title":"Mathematics of Control, Signals and Systems"},{"issue":"1","key":"10.1016\/j.neunet.2026.108559_bib0006","doi-asserted-by":"crossref","first-page":"127","DOI":"10.1007\/s00365-021-09548-z","article-title":"Nonlinear approximation and (deep) ReLU networks","volume":"55","author":"Daubechies","year":"2022","journal-title":"Constructive Approximation"},{"key":"10.1016\/j.neunet.2026.108559_bib0007","unstructured":"E, W., Ma, C., & Wu, L. (2019). Barron spaces and the compositional function spaces for neural network models. arXiv preprint arXiv: 1906.08039."},{"key":"10.1016\/j.neunet.2026.108559_bib0008","doi-asserted-by":"crossref","first-page":"161","DOI":"10.1017\/S0962492914000129","article-title":"Generalized barycentric coordinates and applications","volume":"24","author":"Floater","year":"2015","journal-title":"Acta Numerica"},{"issue":"3","key":"10.1016\/j.neunet.2026.108559_bib0009","doi-asserted-by":"crossref","first-page":"580","DOI":"10.2307\/1968813","article-title":"Simplizialzerlegungen von beschr\u00e4nkter flachheit","volume":"43","author":"Freudenthal","year":"1942","journal-title":"Annals of Mathematics"},{"key":"10.1016\/j.neunet.2026.108559_bib0010","series-title":"Automata, languages and programming: 16th international colloquium stresa, italy, july 11\u201315, 1989 proceedings 16","first-page":"451","article-title":"Tensor rank is NP-complete","author":"H\u00e5stad","year":"1989"},{"issue":"3","key":"10.1016\/j.neunet.2026.108559_bib0011","doi-asserted-by":"crossref","first-page":"502","DOI":"10.4208\/jcm.1901-m2018-0160","article-title":"ReLU deep neural networks and linear finite elements","volume":"38","author":"He","year":"2020","journal-title":"Journal of Computational Mathematics"},{"key":"10.1016\/j.neunet.2026.108559_bib0012","unstructured":"He, J., & Xu, J. (2023). Deep neural networks and finite elements of any order on arbitrary dimensions. arXiv preprint arXiv: 2312.14276."},{"key":"10.1016\/j.neunet.2026.108559_bib0013","first-page":"3336","article-title":"Towards lower bounds on the depth of ReLU neural networks","volume":"34","author":"Hertrich","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"issue":"5","key":"10.1016\/j.neunet.2026.108559_bib0014","doi-asserted-by":"crossref","first-page":"359","DOI":"10.1016\/0893-6080(89)90020-8","article-title":"Multilayer feedforward networks are universal approximators","volume":"2","author":"Hornik","year":"1989","journal-title":"Neural Networks"},{"issue":"6","key":"10.1016\/j.neunet.2026.108559_bib0015","doi-asserted-by":"crossref","first-page":"A3490","DOI":"10.1137\/22M1477751","article-title":"MIONet: Learning multiple-input operators via tensor product","volume":"44","author":"Jin","year":"2022","journal-title":"SIAM Journal on Scientific Computing"},{"key":"10.1016\/j.neunet.2026.108559_bib0016","doi-asserted-by":"crossref","first-page":"3","DOI":"10.1090\/conm\/114\/1097862","article-title":"Some recent results on convex polytopes","volume":"114","author":"Lee","year":"1990","journal-title":"Contemporary Math"},{"key":"10.1016\/j.neunet.2026.108559_bib0017","article-title":"Convergence analysis of two-layer neural networks with ReLU activation","volume":"30","author":"Li","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"issue":"5","key":"10.1016\/j.neunet.2026.108559_bib0018","doi-asserted-by":"crossref","first-page":"1671","DOI":"10.4208\/cicp.OA-2020-0165","article-title":"Dying ReLU and Initialization: Theory and Numerical Examples","volume":"28","author":"Lu","year":"2020","journal-title":"Communications in Computational Physics"},{"issue":"27","key":"10.1016\/j.neunet.2026.108559_bib0019","doi-asserted-by":"crossref","first-page":"296","DOI":"10.1090\/bproc\/186","article-title":"Piecewise linear functions representable with infinite width shallow ReLU neural networks","volume":"10","author":"McCarty","year":"2023","journal-title":"Proceedings of the American Mathematical Society, Series B"},{"key":"10.1016\/j.neunet.2026.108559_bib0020","series-title":"Proceedings of the 27th international conference on machine learning (ICML-10)","first-page":"807","article-title":"Rectified linear units improve restricted boltzmann machines","author":"Nair","year":"2010"},{"key":"10.1016\/j.neunet.2026.108559_bib0021","unstructured":"Pascanu, R., Montufar, G., & Bengio, Y. (2013). On the number of response regions of deep feed forward networks with piece-wise linear activations. arXiv preprint arXiv: 1312.6098."},{"key":"10.1016\/j.neunet.2026.108559_bib0022","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/j.acha.2021.12.005","article-title":"High-order approximation rates for shallow neural networks with cosine and ReLUk activation functions","volume":"58","author":"Siegel","year":"2022","journal-title":"Applied and Computational Harmonic Analysis"},{"issue":"6","key":"10.1016\/j.neunet.2026.108559_bib0023","doi-asserted-by":"crossref","first-page":"1714","DOI":"10.4208\/jcm.2307-m2022-0233","article-title":"Tensor Neural Network and Its Numerical Integration","volume":"42","author":"Wang","year":"2024","journal-title":"Journal of Computational Mathematics"},{"key":"10.1016\/j.neunet.2026.108559_bib0024","doi-asserted-by":"crossref","first-page":"103","DOI":"10.1016\/j.neunet.2017.07.002","article-title":"Error bounds for approximations with deep ReLU networks","volume":"94","author":"Yarotsky","year":"2017","journal-title":"Neural Networks"}],"container-title":["Neural Networks"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026000225?httpAccept=text\/xml","content-type":"text\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026000225?httpAccept=text\/plain","content-type":"text\/plain","content-version":"vor","intended-application":"text-mining"}],"deposited":{"date-parts":[[2026,4,6]],"date-time":"2026-04-06T18:40:03Z","timestamp":1775500803000},"score":1,"resource":{"primary":{"URL":"https:\/\/linkinghub.elsevier.com\/retrieve\/pii\/S0893608026000225"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":24,"alternative-id":["S0893608026000225"],"URL":"https:\/\/doi.org\/10.1016\/j.neunet.2026.108559","relation":{},"ISSN":["0893-6080"],"issn-type":[{"value":"0893-6080","type":"print"}],"subject":[],"published":{"date-parts":[[2026,6]]},"assertion":[{"value":"Elsevier","name":"publisher","label":"This article is maintained by"},{"value":"Two-hidden-layer ReLU neural networks and finite elements","name":"articletitle","label":"Article Title"},{"value":"Neural Networks","name":"journaltitle","label":"Journal Title"},{"value":"https:\/\/doi.org\/10.1016\/j.neunet.2026.108559","name":"articlelink","label":"CrossRef DOI link to publisher maintained version"},{"value":"article","name":"content_type","label":"Content Type"},{"value":"\u00a9 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.","name":"copyright","label":"Copyright"}],"article-number":"108559"}}