{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T19:58:19Z","timestamp":1781726299832,"version":"3.54.5"},"reference-count":45,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"3","license":[{"start":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T00:00:00Z","timestamp":1777593600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T00:00:00Z","timestamp":1777593600000},"content-version":"am","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T00:00:00Z","timestamp":1777593600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T00:00:00Z","timestamp":1777593600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"NSF","award":["2031985"],"award-info":[{"award-number":["2031985"]}]},{"DOI":"10.13039\/100000893","name":"Simons Foundation","doi-asserted-by":"publisher","award":["814201"],"award-info":[{"award-number":["814201"]}],"id":[{"id":"10.13039\/100000893","id-type":"DOI","asserted-by":"publisher"}]},{"name":"ONR MURI","award":["503405-78051"],"award-info":[{"award-number":["503405-78051"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Signal Process. Mag."],"published-print":{"date-parts":[[2026,5]]},"DOI":"10.1109\/msp.2026.3665010","type":"journal-article","created":{"date-parts":[[2026,6,12]],"date-time":"2026-06-12T19:44:54Z","timestamp":1781293494000},"page":"37-51","source":"Crossref","is-referenced-by-count":0,"title":["On the Convergence, Implicit Bias, and Edge of Stability of Gradient Descent in Deep Learning: Reviewing recent progress [Special Issue on the Mathematics of Deep Learning]"],"prefix":"10.1109","volume":"43","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0600-3854","authenticated-orcid":false,"given":"Hancheng","family":"Min","sequence":"first","affiliation":[{"name":"Institute of Natural Sciences (INS), the School of Mathematical Sciences (SMS), and the MOE-LSC, Shanghai Jiao Tong University, Shanghai, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3601-9777","authenticated-orcid":false,"given":"Lachlan Ewen","family":"MacDonald","sequence":"additional","affiliation":[{"name":"Center for Innovation in Data Engineering and Science (IDEAS) and the Department of Electrical and Systems Engineering (ESE), School of Engineering and Applied Science, University of Pennsylvania, Philadelphia, PA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ren\u00e9","family":"Vidal","sequence":"additional","affiliation":[{"name":"Department of Electrical and Systems Engineering and Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"65","article-title":"Linear learning: Landscapes and algorithms","volume-title":"Proc. Advances Neural Inf. Process. Syst. (NeurIPS)","author":"Baldi","year":"1988"},{"key":"ref2","first-page":"1246","article-title":"Gradient descent only converges to minimizers","volume-title":"Proc. 29th Annu. Conf. Learn. Theory","author":"Lee","year":"2016"},{"key":"ref3","first-page":"586","article-title":"Deep learning without poor local minima","volume-title":"Proc. 30th Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Kawaguchi","year":"2016"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2017.467"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1007\/s10208-020-09461-0"},{"key":"ref6","first-page":"8580","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. 32nd Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Jacot","year":"2018"},{"key":"ref7","volume-title":"Introduction to Optimization","author":"Polyak","year":"1987"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1145\/3446776"},{"key":"ref9","first-page":"1675","article-title":"Gradient descent finds global minima of deep neural networks","volume-title":"Proc. Int. Conf. Mach. Learn. (ICML)","author":"Du","year":"2019"},{"key":"ref10","first-page":"8141","article-title":"On exact computation with an infinitely wide neural net","volume-title":"Proc. 33rd Int. Conf. Neural Inf. Process. Syst.","author":"Arora","year":"2019"},{"key":"ref11","first-page":"242","article-title":"A convergence theory for deep learning via over-parameterization","volume-title":"Proc. 36th Int. Conf. Mach. Learn.","author":"Allen-Zhu","year":"2019"},{"key":"ref12","first-page":"8572","article-title":"Wide neural networks of any depth evolve as linear models under gradient descent","volume-title":"Proc. 33rd Int. Conf. Neural Inf. Process. Syst.","author":"Lee","year":"2019"},{"key":"ref13","first-page":"2937","article-title":"On lazy training in differentiable programming","volume-title":"Proc. 33rd Int. Conf. Neural Inf. Process. Syst.","author":"Chizat","year":"2019"},{"key":"ref14","article-title":"Feature learning in infinite-width neural networks","author":"Yang","year":"2021"},{"key":"ref15","first-page":"2388","article-title":"Mean-field theory of two-layers neural networks: Dimension-free bounds and kernel limit","volume-title":"Proc. 32nd Conf. Learn. Theory, in Proceedings of Machine Learning Research","volume":"99","author":"Mei","year":"2019"},{"key":"ref16","first-page":"22182","article-title":"Implicit bias in deep linear classification: Initialization scale vs training accuracy","volume-title":"Proc. 34th Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Moroshko","year":"2020"},{"key":"ref17","first-page":"14,705","article-title":"Language models are few-shot learners","volume-title":"Proc. 37th Int. Conf. Neural Inf. Process. Syst.","author":"MacDonald","year":"2023"},{"key":"ref18","article-title":"A convergence analysis of gradient descent for deep linear neural networks","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Arora","year":"2018"},{"key":"ref19","first-page":"7760","article-title":"On the explicit role of initialization on the convergence and implicit bias of overparametrized linear networks","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Min","year":"2021"},{"key":"ref20","first-page":"63,210","article-title":"Abide by the law and follow the flow: Conservation laws for gradient flows","volume-title":"Proc. 37th Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Marcotte","year":"2023"},{"key":"ref21","first-page":"24,850","article-title":"On the convergence of gradient flow on multi-layer linear models","volume-title":"Proc. 40th Int. Conf. Mach. Learn.","author":"Min","year":"2023"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1093\/imaiai\/iaaa039"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/tit.2006.871582"},{"key":"ref24","first-page":"3635","article-title":"Kernel and rich regimes in overparametrized models","volume-title":"Proc. 33rd Conf. Learn. Theory","author":"Woodworth","year":"2020"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/msp.2018.2821706"},{"key":"ref26","article-title":"A unifying view on implicit bias in training linear neural networks","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Yun","year":"2021"},{"key":"ref27","first-page":"9482","article-title":"Implicit bias of gradient descent on linear convolutional networks","volume-title":"Proc. 32nd Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Gunasekar","year":"2018"},{"key":"ref28","first-page":"6152","article-title":"Implicit regularization in matrix factorization","volume-title":"Proc. 31st Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Gunasekar","year":"2017"},{"key":"ref29","article-title":"Towards resolving the implicit bias of gradient descent for matrix factorization: Greedy low-rank learning","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Li","year":"2021"},{"key":"ref30","first-page":"15,200","article-title":"Understanding incremental learning of gradient descent: A fine-grained analysis of matrix sensing","volume-title":"Proc. 40th Int. Conf. Mach. Learn. (ICML)","author":"Jin","year":"2023"},{"key":"ref31","first-page":"468","article-title":"On the implicit bias of initialization shape: Beyond infinitesimal mirror descent","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Azulay","year":"2021"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1561\/2200000050"},{"key":"ref33","first-page":"7413","article-title":"Implicit regularization in deep matrix factorization","volume-title":"Proc. 33rd Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Arora","year":"2019"},{"key":"ref34","article-title":"Exact solutions to the nonlinear dynamics of learning in deep linear neural networks","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Saxe","year":"2014"},{"key":"ref35","first-page":"24,519","article-title":"Transformers learn through gradual rank increase","volume-title":"Proc. 37th Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Boix-Adsera","year":"2023"},{"key":"ref36","first-page":"23,831","article-title":"Small random initialization is akin to spectral learning: Optimization and generalization guarantees for overparameterized low-rank matrix reconstruction","volume-title":"Proc. 35th Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Stger","year":"2021"},{"key":"ref37","article-title":"Early neuron alignment in two-layer ReLU networks with small initialization","volume-title":"Proc. 12th Int. Conf. Learn. Representations","author":"Min","year":"2024"},{"issue":"70","key":"ref38","first-page":"1","article-title":"The implicit bias of gradient descent on separable data","volume":"19","author":"Soudry","year":"2018","journal-title":"J. Mach. Learn. Res."},{"key":"ref39","first-page":"17,176","article-title":"Directional convergence and alignment in deep learning","volume-title":"Proc. 34th Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"Ji","year":"2020"},{"key":"ref40","article-title":"Gradient descent maximizes the margin of homogeneous neural networks","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Lyu","year":"2020"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1145\/3571070"},{"key":"ref42","article-title":"Self-stabilization: The implicit bias of gradient descent at the edge of stability","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Damian","year":"2023"},{"key":"ref43","article-title":"Gradient descent on neural networks typically occurs at the edge of stability","volume-title":"Proc. Int. Conf. Learn. Representations (ICLR)","author":"Cohen","year":"2021"},{"key":"ref44","first-page":"5019","article-title":"Large stepsize gradient descent for logistic loss: Non-monotonicity of the loss improves optimization efficiency","volume-title":"Proc. 37th Conf. Learn. Theory","author":"Wu","year":"2024"},{"key":"ref45","article-title":"Convergence rates for gradient descent on the edge of stability for overparametrised least squares","volume-title":"Proc. 39th Conf. Neural Inf. Process. Syst. (NeurIPS)","author":"MacDonald","year":"2025"}],"container-title":["IEEE Signal Processing Magazine"],"original-title":[],"link":[{"URL":"https:\/\/ieeexplore.ieee.org\/ielam\/79\/11561017\/11561020-aam.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/79\/11561017\/11561020.pdf?arnumber=11561020","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T19:41:14Z","timestamp":1781725274000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11561020\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5]]},"references-count":45,"journal-issue":{"issue":"3"},"URL":"https:\/\/doi.org\/10.1109\/msp.2026.3665010","relation":{},"ISSN":["1053-5888","1558-0792"],"issn-type":[{"value":"1053-5888","type":"print"},{"value":"1558-0792","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5]]}}}