{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T09:15:06Z","timestamp":1761556506015,"version":"build-2065373602"},"reference-count":47,"publisher":"IOP Publishing","issue":"4","license":[{"start":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T00:00:00Z","timestamp":1761523200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T00:00:00Z","timestamp":1761523200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2025,12,30]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Deep neural networks have achieved exceptional performance across various fields by learning complex, nonlinear mappings from large-scale datasets. However, they face challenges such as high memory requirements and limited interpretability. This paper introduces an approach where master equations of physics are converted into multilayered networks that are trained via backpropagation. The resulting general-purpose model effectively encodes data in the properties of the underlying physical system. In contrast to existing methods wherein a trained neural network is used as a computationally efficient alternative for solving physical equations, our approach directly treats physics equations as trainable models. Rather than approximating physics with a neural network or augmenting a network with physics-inspired constraints, this framework makes the equation itself the architecture. We demonstrate this physical embedding concept with the nonlinear Schr\u00f6dinger equation, which acts as trainable architecture for learning complex patterns including nonlinear mappings and memory effects from data. The network embeds data representation in orders of magnitude fewer parameters than conventional neural networks when tested on time series data. Notably, the trained \u2018Nonlinear Schr\u00f6dinger Network\u2019 is interpretable, with all parameters having physical meanings. Curiously, this approach also provides a blueprint for implementing such AI computations in physical analog systems, offering a direct path toward low-latency and energy-efficient hardware realizations. The proposed method is also extended to the Gross-Pitaevskii equation, demonstrating the broad applicability of the framework to other master equations of physics. Among our results, an ablation study quantifies the relative importance of physical terms such as dispersion, nonlinearity, and potential energy for classification accuracy. We also outline the limitations and benefits of this approach as it relates to universality and generalizability. Overall, this work aims to establish physical embedding as a path toward compact, interpretable AI models, bridging machine learning and fundamental physics.<\/jats:p>","DOI":"10.1088\/2632-2153\/ae0f37","type":"journal-article","created":{"date-parts":[[2025,10,3]],"date-time":"2025-10-03T22:49:37Z","timestamp":1759531777000},"page":"045018","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Physical data embedding for memory efficient AI"],"prefix":"10.1088","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0009-0000-2761-7148","authenticated-orcid":true,"given":"Callen","family":"MacPhee","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yiming","family":"Zhou","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0169-8231","authenticated-orcid":true,"given":"Bahram","family":"Jalali","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"266","published-online":{"date-parts":[[2025,10,27]]},"reference":[{"key":"mlstae0f37bib1","first-page":"pp 1","type":"conference-proceedings","article-title":"Zero-infinity: breaking the gpu memory wall for extreme scale deep learning","author":"Rajbhandari","year":"2021"},{"key":"mlstae0f37bib2","doi-asserted-by":"publisher","first-page":"33","DOI":"10.1109\/MM.2024.3373763","type":"journal-article","article-title":"Ai and memory wall","volume":"44","author":"Gholami","year":"2024","journal-title":"IEEE Micro"},{"key":"mlstae0f37bib3","doi-asserted-by":"publisher","first-page":"300","DOI":"10.1016\/j.jpdc.2022.01.013","type":"journal-article","article-title":"Optically connected memory for disaggregated data centers","volume":"163","author":"Gonzalez","year":"2022","journal-title":"J. Parallel Distrib. Comput."},{"key":"mlstae0f37bib4","doi-asserted-by":"publisher","first-page":"2214","DOI":"10.1109\/JLT.2021.3136570","type":"journal-article","article-title":"Prospects and challenges of photonic switching in data centers and computing systems","volume":"40","author":"Yoo","year":"2021","journal-title":"J. Lightwave Technol."},{"key":"mlstae0f37bib5","doi-asserted-by":"publisher","first-page":"13385","DOI":"10.1007\/s11227-021-03805-5","type":"journal-article","article-title":"A survey of energy-saving technologies in cloud data centers","volume":"77","author":"Cheng","year":"2021","journal-title":"J. Supercomput."},{"key":"mlstae0f37bib6","first-page":"795","type":"journal-article","article-title":"Sustainable AI: environmental implications, challenges and opportunities","volume":"4","author":"Wu","year":"2022","journal-title":"Proc. Mach. Learn. Syst."},{"year":"2020","author":"Molnar","key":"mlstae0f37bib7","type":"book"},{"key":"mlstae0f37bib8","doi-asserted-by":"publisher","first-page":"31","DOI":"10.1145\/3236386.3241340","type":"journal-article","article-title":"The mythos of model interpretability: in machine learning, the concept of interpretability is both important and slippery","volume":"16","author":"Lipton","year":"2018","journal-title":"Queue"},{"key":"mlstae0f37bib9","doi-asserted-by":"publisher","first-page":"206","DOI":"10.1038\/s42256-019-0048-x","type":"journal-article","article-title":"Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead","volume":"1","author":"Rudin","year":"2019","journal-title":"Nat. Mach. Intell."},{"key":"mlstae0f37bib10","doi-asserted-by":"publisher","first-page":"422","DOI":"10.1038\/s42254-021-00314-5","type":"journal-article","article-title":"Physics-informed machine learning","volume":"3","author":"Karniadakis","year":"2021","journal-title":"Nat. Rev. Phys."},{"key":"mlstae0f37bib11","type":"journal-article","article-title":"Physics-AI symbiosis","volume":"3","author":"Jalali","year":"2022","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"mlstae0f37bib12","doi-asserted-by":"publisher","DOI":"10.1103\/RevModPhys.91.045002","type":"journal-article","article-title":"Machine learning and the physical sciences","volume":"91","author":"Carleo","year":"2019","journal-title":"Rev. Mod. Phys."},{"key":"mlstae0f37bib13","doi-asserted-by":"publisher","first-page":"71050","DOI":"10.1109\/ACCESS.2020.2987324","type":"journal-article","article-title":"Driven by data or derived through physics? A review of hybrid physics guided machine learning techniques with cyber-physical system (CPS) focus","volume":"8","author":"Rai","year":"2020","journal-title":"IEEE Access"},{"key":"mlstae0f37bib14","doi-asserted-by":"publisher","first-page":"686","DOI":"10.1016\/j.jcp.2018.10.045","type":"journal-article","article-title":"Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations","volume":"378","author":"Raissi","year":"2019","journal-title":"J. Comput. Phys."},{"year":"2020","author":"Li","article-title":"Fourier neural operator for parametric partial differential equations","key":"mlstae0f37bib15","type":"preprint"},{"key":"mlstae0f37bib16","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevX.9.021032","type":"journal-article","article-title":"Large-scale optical neural networks based on photoelectric multiplication","volume":"9","author":"Hamerly","year":"2019","journal-title":"Phys. Rev. X"},{"key":"mlstae0f37bib17","doi-asserted-by":"publisher","first-page":"1004","DOI":"10.1126\/science.aat8084","type":"journal-article","article-title":"All-optical machine learning using diffractive deep neural networks","volume":"361","author":"Lin","year":"2018","journal-title":"Science"},{"key":"mlstae0f37bib18","doi-asserted-by":"publisher","first-page":"1333","DOI":"10.1126\/science.aaw2498","type":"journal-article","article-title":"Inverse-designed metastructures that solve equations","volume":"363","author":"Mohammadi Estakhri","year":"2019","journal-title":"Science"},{"key":"mlstae0f37bib19","doi-asserted-by":"publisher","first-page":"1308","DOI":"10.1109\/JLT.2022.3146131","type":"journal-article","article-title":"Nonlinear Schr\u00f6dinger kernel for hardware acceleration of machine learning","volume":"40","author":"Zhou","year":"2022","journal-title":"J. Lightwave Technol."},{"key":"mlstae0f37bib20","doi-asserted-by":"publisher","first-page":"108","DOI":"10.1109\/JPROC.2018.2871057","type":"journal-article","article-title":"The next generation of deep learning hardware: analog computing","volume":"107","author":"Haensch","year":"2018","journal-title":"Proc. IEEE"},{"key":"mlstae0f37bib21","doi-asserted-by":"publisher","first-page":"6570","DOI":"10.1021\/acs.nanolett.8b03171","type":"journal-article","article-title":"Generative model for the inverse design of metasurfaces","volume":"18","author":"Liu","year":"2018","journal-title":"Nano Lett."},{"key":"mlstae0f37bib22","doi-asserted-by":"publisher","DOI":"10.1063\/5.0071616","type":"journal-article","article-title":"Maxwellnet: physics-driven deep neural network training based on Maxwell\u2019s equations","volume":"7","author":"Lim","year":"2022","journal-title":"APL Photonics"},{"key":"mlstae0f37bib23","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1017\/jfm.2016.803","type":"journal-article","article-title":"Deep learning in fluid dynamics","volume":"814","author":"Kutz","year":"2017","journal-title":"J. Fluid Mech."},{"key":"mlstae0f37bib24","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s40304-018-0127-z","type":"journal-article","article-title":"The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems","volume":"6","author":"Yu","year":"2018","journal-title":"Commun. Math. Stat."},{"key":"mlstae0f37bib25","doi-asserted-by":"publisher","first-page":"3603","DOI":"10.1021\/acs.jpclett.4c00598","type":"journal-article","article-title":"Artificial-intelligence-based surrogate solution of dissipative quantum dynamics: physics-informed reconstruction of the universal propagator","volume":"15","author":"Zhang","year":"2024","journal-title":"J. Phys. Chem. Lett."},{"key":"mlstae0f37bib26","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevResearch.7.L012013","type":"journal-article","article-title":"Neural quantum propagators for driven-dissipative quantum dynamics","volume":"7","author":"Zhang","year":"2025","journal-title":"Phys. Rev. Res."},{"key":"mlstae0f37bib27","type":"journal-article","article-title":"Neural network representation of quantum systems","volume":"5","author":"Hashimoto","year":"2024","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"mlstae0f37bib28","doi-asserted-by":"publisher","first-page":"218","DOI":"10.1038\/s42256-021-00302-5","type":"journal-article","article-title":"Learning nonlinear operators via deeponet based on the universal approximation theorem of operators","volume":"3","author":"Lu","year":"2021","journal-title":"Nat. Mach. Intell."},{"year":"2018","author":"Gurney","key":"mlstae0f37bib29","type":"book"},{"key":"mlstae0f37bib30","doi-asserted-by":"publisher","first-page":"1798","DOI":"10.1109\/TPAMI.2013.50","type":"journal-article","article-title":"Representation learning: a review and new perspectives","volume":"35","author":"Bengio","year":"2013","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"mlstae0f37bib31","first-page":"pp 195","type":"book","article-title":"Nonlinear fiber optics","author":"Agrawal","year":"2000"},{"key":"mlstae0f37bib32","doi-asserted-by":"publisher","first-page":"1293","DOI":"10.1109\/JAS.2019.1911747","type":"journal-article","article-title":"The UCR time series archive","volume":"6","author":"Dau","year":"2019","journal-title":"IEEE\/CAA J. Autom. Sin."},{"key":"mlstae0f37bib33","doi-asserted-by":"publisher","first-page":"281","DOI":"10.1007\/s10994-008-5093-3","type":"journal-article","article-title":"Finding anomalous periodic time series: an application to catalogs of periodic variable stars","volume":"74","author":"Rebbapragada","year":"2009","journal-title":"Mach. Learn."},{"year":"2018","author":"Jackson","article-title":"Jakobovski\/free-spoken-digit-dataset: v1. 0.8","key":"mlstae0f37bib34","type":"other"},{"key":"mlstae0f37bib35","doi-asserted-by":"publisher","first-page":"917","DOI":"10.1007\/s10618-019-00619-1","type":"journal-article","article-title":"Deep learning for time series classification: a review","volume":"33","author":"Ismail Fawaz","year":"2019","journal-title":"Data Min. Knowl. Discov."},{"year":"2017","author":"Hestness","article-title":"Deep learning scaling is predictable, empirically","key":"mlstae0f37bib36","type":"preprint"},{"year":"2019","author":"Rosenfeld","article-title":"A constructive prediction of the generalization error across scales","key":"mlstae0f37bib37","type":"preprint"},{"key":"mlstae0f37bib38","doi-asserted-by":"publisher","first-page":"318","DOI":"10.1016\/S0021-9991(03)00102-5","type":"journal-article","article-title":"Numerical solution of the Gross\u2013Pitaevskii equation for Bose\u2013Einstein condensation","volume":"187","author":"Bao","year":"2003","journal-title":"J. Comput. Phys."},{"key":"mlstae0f37bib39","doi-asserted-by":"publisher","first-page":"179","DOI":"10.1088\/0305-4470\/39\/12\/L02","type":"journal-article","article-title":"Symbolic calculation in development of algorithms: split-step methods for the Gross\u2013Pitaevskii equation","volume":"39","author":"Javanainen","year":"2006","journal-title":"J. Phys. A: Math. Gen."},{"key":"mlstae0f37bib40","first-page":"pp 95","type":"book","article-title":"Operator splitting","author":"MacNamara","year":"2017"},{"key":"mlstae0f37bib41","doi-asserted-by":"publisher","first-page":"424","DOI":"10.1016\/j.spmi.2014.08.007","type":"journal-article","article-title":"An accurate Fourier splitting scheme for solving the cubic quintic complex Ginzburg\u2013Landau equation","volume":"75","author":"Mohammedi","year":"2014","journal-title":"Superlattices Microstruct."},{"key":"mlstae0f37bib42","doi-asserted-by":"publisher","DOI":"10.1016\/j.optcom.2025.132170","type":"journal-article","article-title":"Numerical solving of dissipative solitons and turing patterns in the Lugiato-Lefever equation using squared-operator iteration method","volume":"591","author":"Cheng","year":"2025","journal-title":"Opt. Commun."},{"key":"mlstae0f37bib43","doi-asserted-by":"publisher","first-page":"503","DOI":"10.1016\/S0898-1221(03)80033-0","type":"journal-article","article-title":"A split-step Fourier method for the complex modified Korteweg-de Vries equation","volume":"45","author":"Muslu","year":"2003","journal-title":"Comput. Math. Appl."},{"key":"mlstae0f37bib44","doi-asserted-by":"publisher","first-page":"704","DOI":"10.1038\/nphoton.2015.208","type":"journal-article","article-title":"Analog optical computing","volume":"9","author":"Solli","year":"2015","journal-title":"Nat. Photon."},{"key":"mlstae0f37bib45","doi-asserted-by":"crossref","DOI":"10.1088\/2515-7647\/acff54","type":"journal-article","article-title":"Low latency computing for time stretch instruments","volume":"5","author":"Zhou","year":"2023","journal-title":"J. Phys. Photon."},{"key":"mlstae0f37bib46","doi-asserted-by":"publisher","first-page":"24","DOI":"10.1186\/s43593-022-00034-y","type":"journal-article","article-title":"Vevid: vision enhancement via virtual diffraction and coherent detection","volume":"2","author":"Jalali","year":"2022","journal-title":"eLight"},{"year":"2024","author":"Zhou","article-title":"Nonlinear Schr\u00f6dinger network","key":"mlstae0f37bib47","type":"preprint"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T09:11:38Z","timestamp":1761556298000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae0f37"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,27]]},"references-count":47,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2025,10,27]]},"published-print":{"date-parts":[[2025,12,30]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/ae0f37","relation":{},"ISSN":["2632-2153"],"issn-type":[{"type":"electronic","value":"2632-2153"}],"subject":[],"published":{"date-parts":[[2025,10,27]]},"assertion":[{"value":"Physical data embedding for memory efficient AI","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2025 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2025-04-03","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-10-03","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-10-27","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}