{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,2]],"date-time":"2025-08-02T18:44:42Z","timestamp":1754160282252,"version":"3.41.2"},"reference-count":46,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"8","license":[{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001502","name":"Department of Atomic Energy, Government of India","doi-asserted-by":"publisher","award":["RTI4001"],"award-info":[{"award-number":["RTI4001"]}],"id":[{"id":"10.13039\/501100001502","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Piyush Srivastava was supported in part by Adobe Systems Inc. via a gift to TIFR, from Science and Engineering Research Board, Department of Science and Technology (DST), India","award":["MTR\/2023\/001547"],"award-info":[{"award-number":["MTR\/2023\/001547"]}]},{"name":"Infosys-Chandrasekharan Virtual Centre for Random Geometry at Tata Institute of Fundamental Research"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Inform. Theory"],"published-print":{"date-parts":[[2025,8]]},"DOI":"10.1109\/tit.2025.3567076","type":"journal-article","created":{"date-parts":[[2025,5,5]],"date-time":"2025-05-05T13:55:02Z","timestamp":1746453302000},"page":"6340-6352","source":"Crossref","is-referenced-by-count":0,"title":["A Direct Proof of a Unified Law of Robustness for Bregman Divergence Losses"],"prefix":"10.1109","volume":"71","author":[{"given":"Santanu","family":"Das","sequence":"first","affiliation":[{"name":"Tata Institute of Fundamental Research, Mumbai, India"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7174-9778","authenticated-orcid":false,"given":"Jatin","family":"Batra","sequence":"additional","affiliation":[{"name":"Tata Institute of Fundamental Research, Mumbai, India"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0953-2890","authenticated-orcid":false,"given":"Piyush","family":"Srivastava","sequence":"additional","affiliation":[{"name":"Tata Institute of Fundamental Research, Mumbai, India"}]}],"member":"263","reference":[{"key":"ref1","first-page":"5393","article-title":"Dynamical isometry and a mean field theory of CNNs: How to train 10,000-layer vanilla convolutional neural networks","volume-title":"Proc. 35th Int. Conf. Mach. Learn.","volume":"80","author":"Xiao"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2016.90"},{"key":"ref3","first-page":"1097","article-title":"ImageNet classification with deep convolutional neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst. (NeurIPS)","volume":"25","author":"Krizhevsky"},{"key":"ref4","first-page":"30016","article-title":"Training compute-optimal large language models","volume-title":"Proc. 36th Int. Conf. Neural Inf. Process. Syst.","author":"Hoffmann"},{"key":"ref5","first-page":"28811","article-title":"A universal law of robustness via isoperimetry","volume-title":"Proc. Adv. Neural Inf. Process. Syst. (NeurIPS)","volume":"34","author":"Bubeck"},{"key":"ref6","first-page":"10767","article-title":"Rethinking bias-variance trade-off for generalization of neural networks","volume-title":"Proc. 37th Int. Conf. Mach. Learn. (ICML)","author":"Yang"},{"key":"ref7","first-page":"11856","article-title":"Double-descent curves in neural networks: A new perspective using Gaussian processes","volume-title":"Proc. AAAI Conf. Artif. Intell.","author":"Harzli"},{"article-title":"In search of the real inductive bias: On the role of implicit regularization in deep learning","volume-title":"Proc. 3rd Int. Conf. Learn. Represent. (Workshop Track)","author":"Neyshabur","key":"ref8"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-40994-3_25"},{"article-title":"Intriguing properties of neural networks","volume-title":"2nd Int. Conf. Learn. Represent. (ICLR)","author":"Szegedy","key":"ref10"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2015.7298640"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/spw.2019.00033"},{"key":"ref13","first-page":"125","article-title":"Adversarial examples are not bugs, they are features","volume-title":"Proc. Adv. Neural Inf. Process. Syst. (NeurIPS)","author":"Ilyas"},{"article-title":"Theoretical understanding of learning from adversarial perturbations","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Kumano","key":"ref14"},{"key":"ref15","first-page":"4138","article-title":"Towards deep learning models resistant to adversarial attacks","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"M\u0105dry"},{"key":"ref16","first-page":"541","article-title":"To understand deep learning we need to understand kernel learning","volume-title":"Proc. 35th Int. Conf. Mach. Learn.","author":"Belkin"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-20212-4"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1214\/aop\/1176991418"},{"issue":"125","key":"ref19","first-page":"1","article-title":"Classification with deep neural networks and logistic loss","volume":"25","author":"Zhang","year":"2024","journal-title":"J. Mach. Learn. Res."},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/lgrs.2022.3197617"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1017\/cbo9780511813658"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/tit.2005.850145"},{"key":"ref23","doi-asserted-by":"crossref","DOI":"10.1017\/CBO9781107298019","volume-title":"Understanding Machine Learning: From Theory to Algorithms","author":"Shalev-Shwartz","year":"2014"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1017\/9781108627771"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46379-7_1"},{"key":"ref26","first-page":"463","article-title":"Rademacher and Gaussian complexities: Risk bounds and structural results","volume":"3","author":"Bartlett","year":"2003","journal-title":"J. Mach. Learn. Res."},{"key":"ref27","first-page":"21476","article-title":"On the loss landscape of adversarial training: Identifying challenges and how to overcome them","volume-title":"Proc. 34th Int. Conf. Neural Inf. Process. Syst.","author":"Liu"},{"key":"ref28","first-page":"278","article-title":"Adversarial machine learning at scale","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Kurakin"},{"key":"ref29","first-page":"4799","article-title":"Intriguing properties of adversarial training at scale","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Xie"},{"key":"ref30","first-page":"804","article-title":"A law of robustness for two-layers neural networks","volume-title":"Proc. Conf. Learn. Theory","author":"Bubeck"},{"key":"ref31","first-page":"37439","article-title":"A law of robustness beyond isoperimetry","volume-title":"Proc. 40th Int. Conf. Mach. Learn.","author":"Wu"},{"key":"ref32","first-page":"36094","article-title":"Robustness in deep learning: The good (width), the bad (depth), and the ugly (initialization)","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhu"},{"key":"ref33","first-page":"13029","article-title":"Convergence of adversarial training in overparametrized neural networks","author":"Gao","year":"2019","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"ref34","first-page":"679","article-title":"Over-parameterized adversarial training: An analysis overcoming the curse of dimensionality","volume-title":"Proc. 34th Int. Conf. Neural Inf. Process. Syst. (NeurIPS)","volume":"33","author":"Zhang"},{"key":"ref35","first-page":"2178","article-title":"Generalised Lipschitz regularisation equals distributional robustness","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Cranko"},{"key":"ref36","first-page":"4370","article-title":"Why robust generalization in deep learning is difficult: Perspective of expressive power","volume-title":"Proc. 36th Int. Conf. Neural Inf. Process. Syst.","volume":"35","author":"Li"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1145\/3648351"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1214\/24-aos2353"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1016\/0041-5553(67)90040-7"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1137\/1119012"},{"issue":"4","key":"ref41","doi-asserted-by":"crossref","first-page":"473","DOI":"10.1016\/0047-259X(76)90001-4","article-title":"Exponential inequalities for sums of random vectors","volume":"6","author":"Yurinski\u012d","year":"1976","journal-title":"J. Multivariate Anal."},{"article-title":"A large-deviation inequality for vector-valued martingales","year":"2001","author":"Hayes","key":"ref42"},{"issue":"3","key":"ref43","doi-asserted-by":"crossref","first-page":"1548","DOI":"10.1109\/TIT.2011.2104999","article-title":"Recovering low-rank matrices from few coefficients in any basis","volume":"57","author":"Gross","year":"2011","journal-title":"IEEE Trans. Inf. Theory"},{"key":"ref44","first-page":"22745","article-title":"Training certifiably robust neural networks with efficient local Lipschitz bounds","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Huang"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1137\/22m1478835"},{"key":"ref46","first-page":"4313","article-title":"Gradient descent with early stopping is provably robust to label noise for overparameterized neural networks","volume-title":"Proc. Int. Conf. Artif. Intell. Statist.","author":"Li"}],"container-title":["IEEE Transactions on Information Theory"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/18\/11095909\/10988591.pdf?arnumber=10988591","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,26]],"date-time":"2025-07-26T06:41:05Z","timestamp":1753512065000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10988591\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,8]]},"references-count":46,"journal-issue":{"issue":"8"},"URL":"https:\/\/doi.org\/10.1109\/tit.2025.3567076","relation":{},"ISSN":["0018-9448","1557-9654"],"issn-type":[{"type":"print","value":"0018-9448"},{"type":"electronic","value":"1557-9654"}],"subject":[],"published":{"date-parts":[[2025,8]]}}}