{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,3]],"date-time":"2026-04-03T05:31:37Z","timestamp":1775194297104,"version":"3.50.1"},"reference-count":104,"publisher":"Springer Science and Business Media LLC","issue":"8079","license":[{"start":{"date-parts":[[2025,9,3]],"date-time":"2025-09-03T00:00:00Z","timestamp":1756857600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.springernature.com\/gp\/researchers\/text-and-data-mining"},{"start":{"date-parts":[[2025,9,3]],"date-time":"2025-09-03T00:00:00Z","timestamp":1756857600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.springernature.com\/gp\/researchers\/text-and-data-mining"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Nature"],"published-print":{"date-parts":[[2025,9,4]]},"DOI":"10.1038\/s41586-025-09384-2","type":"journal-article","created":{"date-parts":[[2025,9,3]],"date-time":"2025-09-03T15:02:28Z","timestamp":1756911748000},"page":"53-61","update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":37,"title":["Training of physical neural networks"],"prefix":"10.1038","volume":"645","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-9707-8290","authenticated-orcid":false,"given":"Ali","family":"Momeni","sequence":"first","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0856-9635","authenticated-orcid":false,"given":"Babak","family":"Rahmani","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2407-7470","authenticated-orcid":false,"given":"Benjamin","family":"Scellier","sequence":"additional","affiliation":[]},{"given":"Logan G.","family":"Wright","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1177-9887","authenticated-orcid":false,"given":"Peter L.","family":"McMahon","sequence":"additional","affiliation":[]},{"given":"Clara C.","family":"Wanjura","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9364-4125","authenticated-orcid":false,"given":"Yuhang","family":"Li","sequence":"additional","affiliation":[]},{"given":"Anas","family":"Skalli","sequence":"additional","affiliation":[]},{"given":"Natalia G.","family":"Berloff","sequence":"additional","affiliation":[]},{"given":"Tatsuhiro","family":"Onodera","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2674-684X","authenticated-orcid":false,"given":"Ilker","family":"Oguz","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5858-2811","authenticated-orcid":false,"given":"Francesco","family":"Morichetti","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4821-3924","authenticated-orcid":false,"given":"Philipp","family":"del Hougne","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1600-6151","authenticated-orcid":false,"given":"Manuel","family":"Le Gallo","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5603-5243","authenticated-orcid":false,"given":"Abu","family":"Sebastian","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2440-0944","authenticated-orcid":false,"given":"Azalia","family":"Mirhoseini","sequence":"additional","affiliation":[]},{"given":"Cheng","family":"Zhang","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7521-217X","authenticated-orcid":false,"given":"Danijela","family":"Markovi\u0107","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4003-3056","authenticated-orcid":false,"given":"Daniel","family":"Brunner","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2078-0273","authenticated-orcid":false,"given":"Christophe","family":"Moser","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9914-6231","authenticated-orcid":false,"given":"Sylvain","family":"Gigan","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4566-1753","authenticated-orcid":false,"given":"Florian","family":"Marquardt","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0717-683X","authenticated-orcid":false,"given":"Aydogan","family":"Ozcan","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4866-4490","authenticated-orcid":false,"given":"Julie","family":"Grollier","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2295-2729","authenticated-orcid":false,"given":"Andrea J.","family":"Liu","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4684-8800","authenticated-orcid":false,"given":"Demetri","family":"Psaltis","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4297-5274","authenticated-orcid":false,"given":"Andrea","family":"Al\u00f9","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9486-6854","authenticated-orcid":false,"given":"Romain","family":"Fleury","sequence":"additional","affiliation":[]}],"member":"297","published-online":{"date-parts":[[2025,9,3]]},"reference":[{"key":"9384_CR1","unstructured":"Samborska, V. Scaling up: how increasing inputs has made artificial intelligence more capable. Our World in Data https:\/\/ourworldindata.org\/scaling-up-ai (2025)."},{"key":"9384_CR2","doi-asserted-by":"publisher","first-page":"529","DOI":"10.1038\/s41565-020-0655-z","volume":"15","author":"A Sebastian","year":"2020","unstructured":"Sebastian, A., Le Gallo, M., Khaddam-Aljameh, R. & Eleftheriou, E. Memory devices and applications for in-memory computing. Nat. Nanotechnol. 15, 529\u2013544 (2020).","journal-title":"Nat. Nanotechnol."},{"key":"9384_CR3","doi-asserted-by":"publisher","first-page":"39","DOI":"10.1038\/s41586-020-2973-6","volume":"588","author":"G Wetzstein","year":"2020","unstructured":"Wetzstein, G. et al. Inference in artificial intelligence with deep optics and photonics. Nature 588, 39\u201347 (2020).","journal-title":"Nature"},{"key":"9384_CR4","doi-asserted-by":"publisher","first-page":"549","DOI":"10.1038\/s41586-021-04223-6","volume":"601","author":"LG Wright","year":"2022","unstructured":"Wright, L. G. et al. Deep physical neural networks trained with backpropagation. Nature 601, 549\u2013555 (2022).","journal-title":"Nature"},{"key":"9384_CR5","doi-asserted-by":"publisher","first-page":"100","DOI":"10.1016\/j.neunet.2019.03.005","volume":"115","author":"G Tanaka","year":"2019","unstructured":"Tanaka, G. et al. Recent advances in physical reservoir computing: a review. Neural Netw. 115, 100\u2013123 (2019).","journal-title":"Neural Netw."},{"key":"9384_CR6","doi-asserted-by":"publisher","first-page":"eaay6946","DOI":"10.1126\/sciadv.aay6946","volume":"5","author":"TW Hughes","year":"2019","unstructured":"Hughes, T. W., Williamson, I. A., Minkov, M. & Fan, S. Wave physics as an analog recurrent neural network. Sci. Adv. 5, eaay6946 (2019).","journal-title":"Sci. Adv."},{"key":"9384_CR7","unstructured":"Onodera, T. et al. Scaling on-chip photonic neural processors using arbitrarily programmable wave propagation. Preprint at https:\/\/arxiv.org\/abs\/2402.17750 (2024)."},{"key":"9384_CR8","doi-asserted-by":"publisher","first-page":"1297","DOI":"10.1126\/science.adi8474","volume":"382","author":"A Momeni","year":"2023","unstructured":"Momeni, A., Rahmani, B., Mall\u00e9jac, M., del Hougne, P. & Fleury, R. Backpropagation-free training of deep physical neural networks. Science 382, 1297\u20131303 (2023).","journal-title":"Science"},{"key":"9384_CR9","doi-asserted-by":"publisher","first-page":"202","DOI":"10.1126\/science.adl1203","volume":"384","author":"Z Xu","year":"2024","unstructured":"Xu, Z. et al. Large-scale photonic chiplet Taichi empowers 160-TOPS\/W artificial general intelligence. Science 384, 202\u2013209 (2024).","journal-title":"Science"},{"key":"9384_CR10","doi-asserted-by":"publisher","first-page":"533","DOI":"10.1038\/323533a0","volume":"323","author":"DE Rumelhart","year":"1986","unstructured":"Rumelhart, D. E., Hinton, G. E. & Williams, R. J. Learning representations by back-propagating errors. Nature 323, 533\u2013536 (1986).","journal-title":"Nature"},{"key":"9384_CR11","doi-asserted-by":"publisher","first-page":"1004","DOI":"10.1126\/science.aat8084","volume":"361","author":"X Lin","year":"2018","unstructured":"Lin, X. et al. All-optical machine learning using diffractive deep neural networks. Science 361, 1004\u20131008 (2018).","journal-title":"Science"},{"key":"9384_CR12","doi-asserted-by":"publisher","first-page":"680","DOI":"10.1038\/s41928-023-01010-1","volume":"6","author":"M Le Gallo","year":"2023","unstructured":"Le Gallo, M. et al. A 64-core mixed-signal in-memory compute chip based on phase-change memory for deep neural network inference. Nat. Electron. 6, 680\u2013693 (2023).","journal-title":"Nat. Electron."},{"key":"9384_CR13","doi-asserted-by":"publisher","first-page":"723","DOI":"10.1038\/s41566-023-01233-w","volume":"17","author":"Z Chen","year":"2023","unstructured":"Chen, Z. et al. Deep learning with coherent VCSEL neural networks. Nat. Photon. 17, 723\u2013730 (2023).","journal-title":"Nat. Photon."},{"key":"9384_CR14","doi-asserted-by":"publisher","first-page":"4207","DOI":"10.1515\/nanoph-2020-0291","volume":"9","author":"D Mengu","year":"2020","unstructured":"Mengu, D. et al. Misalignment resilient diffractive optical networks. Nanophotonics 9, 4207\u20134219 (2020).","journal-title":"Nanophotonics"},{"key":"9384_CR15","doi-asserted-by":"publisher","first-page":"19662","DOI":"10.1364\/OE.17.019662","volume":"17","author":"K Matsushima","year":"2009","unstructured":"Matsushima, K. & Shimobaba, T. Band-limited angular spectrum method for numerical simulation of free-space propagation in far and near fields. Opt. Express 17, 19662\u201319673 (2009).","journal-title":"Opt. Express"},{"key":"9384_CR16","first-page":"9346","volume":"33","author":"J Launay","year":"2020","unstructured":"Launay, J., Poli, I., Boniface, F. & Krzakala, F. Direct feedback alignment scales to modern deep learning tasks and architectures. Adv. Neural Inf. Process. Syst. 33, 9346\u20139360 (2020).","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"9384_CR17","doi-asserted-by":"publisher","first-page":"e2109194119","DOI":"10.1073\/pnas.2109194119","volume":"119","author":"B Cramer","year":"2022","unstructured":"Cramer, B. et al. Surrogate gradients for analog neuromorphic computing. Proc. Natl Acad. Sci. 119, e2109194119 (2022).","journal-title":"Proc. Natl Acad. Sci."},{"key":"9384_CR18","doi-asserted-by":"publisher","first-page":"803","DOI":"10.1364\/OPTICA.456108","volume":"9","author":"J Spall","year":"2022","unstructured":"Spall, J., Guo, X. & Lvovsky, A. I. Hybrid training of optical neural networks. Optica 9, 803\u2013811 (2022).","journal-title":"Optica"},{"key":"9384_CR19","doi-asserted-by":"publisher","DOI":"10.1038\/ncomms13276","volume":"7","author":"TP Lillicrap","year":"2016","unstructured":"Lillicrap, T. P., Cownden, D., Tweed, D. B. & Akerman, C. J. Random synaptic feedback weights support error backpropagation for deep learning. Nat. Commun. 7, 13276 (2016).","journal-title":"Nat. Commun."},{"key":"9384_CR20","doi-asserted-by":"crossref","unstructured":"Brunton, S. L. & Kutz, J. N. Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control (Cambridge Univ. Press, 2022).","DOI":"10.1017\/9781009089517"},{"key":"9384_CR21","unstructured":"Hinton, G. The forward-forward algorithm: some preliminary investigations. Preprint at https:\/\/arxiv.org\/abs\/2212.13345 (2022)."},{"key":"9384_CR22","unstructured":"Laydevant, J., Lott, A., Venturelli, D. & McMahon, P. L. The benefits of self-supervised learning for training physical neural networks. In Proc. 37th First Workshop on Machine Learning with New Compute Paradigms at NeurIPS 2023 (MLNPCP 2023) https:\/\/openreview.net\/forum?id=Fik4cO7FXd (OpenReview, 2023)."},{"key":"9384_CR23","unstructured":"Refinetti, M., d\u2019Ascoli, S., Ohana, R. & Goldt, S. Align, then memorise: the dynamics of learning with feedback alignment. In Proc. 38th International Conference on Machine Learning, 8925\u20138935 (MLR Press, 2021)."},{"key":"9384_CR24","unstructured":"Lillicrap, T. P., Cownden, D., Tweed, D. B. & Akerman, C. J. Random feedback weights support learning in deep neural networks. Preprint at https:\/\/arxiv.org\/abs\/1411.0247 (2014)."},{"key":"9384_CR25","unstructured":"Launay, J. et al. Hardware beyond backpropagation: a photonic co-processor for direct feedback alignment. Preprint at https:\/\/arxiv.org\/abs\/2012.06373 (2020)."},{"key":"9384_CR26","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-022-35216-2","volume":"13","author":"M Nakajima","year":"2022","unstructured":"Nakajima, M. et al. Physical deep learning with biologically inspired training method: gradient-free approach for physical hardware. Nat. Commun. 13, 7847 (2022).","journal-title":"Nat. Commun."},{"key":"9384_CR27","doi-asserted-by":"publisher","first-page":"1158","DOI":"10.1126\/science.7761831","volume":"268","author":"GE Hinton","year":"1995","unstructured":"Hinton, G. E., Dayan, P., Frey, B. J. & Neal, R. M. The \u201cwake-sleep\u201d algorithm for unsupervised neural networks. Science 268, 1158\u20131161 (1995).","journal-title":"Science"},{"key":"9384_CR28","unstructured":"L\u00f6we, S., O\u2019Connor, P. & Veeling, B. Putting an end to end-to-end: gradient-isolated learning of representations. In Proc. Advances in Neural Information Processing Systems 32 (NeuroIPS 2019), 3039\u20133051 (ACM, 2019)."},{"key":"9384_CR29","unstructured":"N\u00f8kland, A. & Eidnes, L. H. Training neural networks with local error signals. In Proc. 36th International Conference on Machine Learning, 4839\u20134850 (MLR Press, 2019)."},{"key":"9384_CR30","unstructured":"Siddiqui, S. A., Krueger, D., LeCun, Y. & Deny, S. Blockwise self-supervised learning at scale. Preprint at https:\/\/arxiv.org\/abs\/2302.01647v1 (2023)."},{"key":"9384_CR31","doi-asserted-by":"publisher","first-page":"5249","DOI":"10.1364\/OL.496884","volume":"48","author":"I Oguz","year":"2023","unstructured":"Oguz, I. et al. Forward\u2013forward training of an optical neural network. Opt. Lett. 48, 5249\u20135252 (2023).","journal-title":"Opt. Lett."},{"key":"9384_CR32","doi-asserted-by":"publisher","first-page":"280","DOI":"10.1038\/s41586-024-07687-4","volume":"632","author":"Z Xue","year":"2024","unstructured":"Xue, Z. et al. Fully forward mode training for optical neural networks. Nature 632, 280\u2013286 (2024).","journal-title":"Nature"},{"key":"9384_CR33","doi-asserted-by":"publisher","first-page":"332","DOI":"10.1109\/9.119632","volume":"37","author":"JC Spall","year":"1992","unstructured":"Spall, J. C. Multivariate stochastic approximation using a simultaneous perturbation gradient approximation. IEEE Trans. Autom. Control 37, 332\u2013341 (1992).","journal-title":"IEEE Trans. Autom. Control"},{"key":"9384_CR34","doi-asserted-by":"publisher","first-page":"026118","DOI":"10.1063\/5.0157645","volume":"1","author":"AN McCaughan","year":"2023","unstructured":"McCaughan, A. N. et al. Multiplexed gradient descent: fast online training of modern datasets on hardware neural networks without backpropagation. APL Mach. Learn. 1, 026118 (2023).","journal-title":"APL Mach. Learn."},{"key":"9384_CR35","doi-asserted-by":"publisher","first-page":"1335","DOI":"10.1038\/s41566-024-01567-z","volume":"18","author":"S Bandyopadhyay","year":"2024","unstructured":"Bandyopadhyay, S. et al. Single-chip photonic deep neural network with forward-only training. Nat. Photon. 18, 1335\u20131343 (2024).","journal-title":"Nat. Photon."},{"key":"9384_CR36","doi-asserted-by":"publisher","first-page":"016002","DOI":"10.1117\/1.AP.6.1.016002","volume":"6","author":"I Oguz","year":"2024","unstructured":"Oguz, I. et al. Programming nonlinear propagation for efficient optical learning machines. Adv. Photonics 6, 016002 (2024).","journal-title":"Adv. Photonics"},{"key":"9384_CR37","doi-asserted-by":"publisher","first-page":"68","DOI":"10.1038\/s42005-025-01972-y","volume":"8","author":"A Skalli","year":"2025","unstructured":"Skalli, A. et al. Annealing-inspired training of an optical neural network with ternary weights. Commun. Phys. 8, 68 (2025).","journal-title":"Commun. Phys."},{"key":"9384_CR38","doi-asserted-by":"publisher","first-page":"756","DOI":"10.1364\/OPTICA.5.000756","volume":"5","author":"J Bueno","year":"2018","unstructured":"Bueno, J. et al. Reinforcement learning in a large-scale photonic recurrent neural network. Optica 5, 756\u2013760 (2018).","journal-title":"Optica"},{"key":"9384_CR39","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-020-66441-8","volume":"10","author":"K Kanno","year":"2020","unstructured":"Kanno, K., Naruse, M. & Uchida, A. Adaptive model selection in photonic reservoir computing by reinforcement learning. Sci. Rep. 10, 10062 (2020).","journal-title":"Sci. Rep."},{"key":"9384_CR40","doi-asserted-by":"publisher","DOI":"10.1038\/ncomms7729","volume":"6","author":"M Hermans","year":"2015","unstructured":"Hermans, M., Burm, M., Van Vaerenbergh, T., Dambre, J. & Bienstman, P. Trainable hardware for dynamical computing using error backpropagation through physical media. Nat. Commun. 6, 6729 (2015).","journal-title":"Nat. Commun."},{"key":"9384_CR41","first-page":"034092","volume":"2","author":"GW Burr","year":"2017","unstructured":"Burr, G. W. et al. Neuromorphic computing using non-volatile memory. Adv. Phys. X 2, 034092 (2017).","journal-title":"Adv. Phys. X"},{"key":"9384_CR42","doi-asserted-by":"publisher","first-page":"398","DOI":"10.1126\/science.ade8450","volume":"380","author":"S Pai","year":"2023","unstructured":"Pai, S. et al. Experimentally realized in situ backpropagation for deep learning in photonic neural networks. Science 380, 398\u2013404 (2023).","journal-title":"Science"},{"key":"9384_CR43","doi-asserted-by":"publisher","first-page":"292","DOI":"10.1109\/JSTQE.2014.2300046","volume":"20","author":"F Morichetti","year":"2014","unstructured":"Morichetti, F. et al. Non-invasive on-chip light observation by contactless waveguide conductivity monitoring. IEEE J. Sel. Top. Quantum Electron. 20, 292\u2013301 (2014).","journal-title":"IEEE J. Sel. Top. Quantum Electron."},{"key":"9384_CR44","doi-asserted-by":"publisher","first-page":"940","DOI":"10.1364\/PRJ.389553","volume":"8","author":"T Zhou","year":"2020","unstructured":"Zhou, T. et al. In situ optical backpropagation training of diffractive optical neural networks. Photonics Res. 8, 940\u2013953 (2020).","journal-title":"Photonics Res."},{"key":"9384_CR45","doi-asserted-by":"publisher","first-page":"B71","DOI":"10.1364\/PRJ.411104","volume":"9","author":"X Guo","year":"2021","unstructured":"Guo, X., Barrett, T. D., Wang, Z. M. & Lvovsky, A. Backpropagation through nonlinear units for the all-optical training of neural networks. Photonics Res. 9, B71\u2013B80 (2021).","journal-title":"Photonics Res."},{"key":"9384_CR46","doi-asserted-by":"publisher","first-page":"1434","DOI":"10.1038\/s41567-024-02534-9","volume":"20","author":"CC Wanjura","year":"2024","unstructured":"Wanjura, C. C. & Marquardt, F. Fully nonlinear neuromorphic computing with linear wave scattering. Nat. Phys. 20, 1434\u20131440 (2024).","journal-title":"Nat. Phys."},{"key":"9384_CR47","doi-asserted-by":"publisher","first-page":"1076","DOI":"10.1038\/s41566-024-01494-z","volume":"18","author":"M Yildirim","year":"2024","unstructured":"Yildirim, M., Dinc, N. U., Oguz, I., Psaltis, D. & Moser, C. Nonlinear processing with linear optics. Nat. Photon. 18, 1076\u20131082 (2024).","journal-title":"Nat. Photon."},{"key":"9384_CR48","doi-asserted-by":"crossref","unstructured":"Xia, F. et al. Nonlinear optical encoding enabled by recurrent linear scattering. Nat. Photon. 18, 1067\u20131075 (2024).","DOI":"10.1038\/s41566-024-01493-0"},{"key":"9384_CR49","doi-asserted-by":"publisher","first-page":"24","DOI":"10.3389\/fncom.2017.00024","volume":"11","author":"B Scellier","year":"2017","unstructured":"Scellier, B. & Bengio, Y. Equilibrium propagation: bridging the gap between energy-based models and backpropagation. Front. Comput. Neurosci. 11, 24 (2017).","journal-title":"Front. Comput. Neurosci."},{"key":"9384_CR50","first-page":"147","volume":"9","author":"DH Ackley","year":"1985","unstructured":"Ackley, D. H., Hinton, G. E. & Sejnowski, T. J. A learning algorithm for Boltzmann machines. Cogn. Sci. 9, 147\u2013169 (1985).","journal-title":"Cogn. Sci."},{"key":"9384_CR51","first-page":"021045","volume":"11","author":"M Stern","year":"2021","unstructured":"Stern, M., Hexner, D., Rocks, J. W. & Liu, A. J. Supervised learning in physical networks: from machine learning to learning machines. Phys. Rev. X 11, 021045 (2021).","journal-title":"Phys. Rev. X"},{"key":"9384_CR52","unstructured":"Scellier, B., Ernoult, M., Kendall, J. & Kumar, S. Energy-based learning algorithms for analog computing: a comparative study. In Proc. 37th International Conference on Neural Information Processing Systems (NIPS \u201923), 52705\u201352731 (ACM, 2023)."},{"key":"9384_CR53","unstructured":"Kendall, J., Pantone, R., Manickavasagam, K., Bengio, Y. & Scellier, B. Training end-to-end analog neural networks with equilibrium propagation. Preprint at https:\/\/arxiv.org\/abs\/2006.01981 (2020)."},{"key":"9384_CR54","doi-asserted-by":"publisher","first-page":"034014","DOI":"10.1088\/2634-4386\/ad752b","volume":"4","author":"Q Wang","year":"2024","unstructured":"Wang, Q., Wanjura, C. C. & Marquardt, F. Training coupled phase oscillators as a neuromorphic platform using equilibrium propagation. Neuromorph. Comput. Eng. 4, 034014 (2024).","journal-title":"Neuromorph. Comput. Eng."},{"key":"9384_CR55","first-page":"45","volume":"6","author":"S-i Yi","year":"2023","unstructured":"Yi, S.-i, Kendall, J. D., Williams, R. S. & Kumar, S. Activity-difference training of deep neural networks using memristor crossbars. Nat. Electron. 6, 45\u201351 (2023).","journal-title":"Nat. Electron."},{"key":"9384_CR56","doi-asserted-by":"publisher","first-page":"3671","DOI":"10.1038\/s41467-024-46879-4","volume":"15","author":"J Laydevant","year":"2024","unstructured":"Laydevant, J., Markovi\u0107, D. & Grollier, J. Training an Ising machine with equilibrium propagation. Nat. Commun. 15, 3671 (2024).","journal-title":"Nat. Commun."},{"key":"9384_CR57","doi-asserted-by":"publisher","first-page":"024053","DOI":"10.1103\/PhysRevApplied.22.024053","volume":"22","author":"LE Altman","year":"2024","unstructured":"Altman, L. E., Stern, M., Liu, A. J. & Durian, D. J. Experimental demonstration of coupled learning in elastic networks. Phys. Rev. Appl. 22, 024053 (2024).","journal-title":"Phys. Rev. Appl."},{"key":"9384_CR58","doi-asserted-by":"publisher","first-page":"014040","DOI":"10.1103\/PhysRevApplied.18.014040","volume":"18","author":"S Dillavou","year":"2022","unstructured":"Dillavou, S., Stern, M., Liu, A. J. & Durian, D. J. Demonstration of decentralized physics-driven learning. Phys. Rev. Appl. 18, 014040 (2022).","journal-title":"Phys. Rev. Appl."},{"key":"9384_CR59","doi-asserted-by":"crossref","unstructured":"Dillavou, S. et al. Machine learning without a processor: emergent learning in a nonlinear analog network. Proc. Natl Acad. Sci. 121, e2319718121 (2024).","DOI":"10.1073\/pnas.2319718121"},{"key":"9384_CR60","doi-asserted-by":"publisher","first-page":"016114","DOI":"10.1063\/5.0181382","volume":"2","author":"M Stern","year":"2024","unstructured":"Stern, M., Dillavou, S., Jayaraman, D., Duria, D. J. & Liu, A. J. Training self-learning circuits for power-efficient solutions. APL Mach. Learn. 2, 016114 (2024).","journal-title":"APL Mach. Learn."},{"key":"9384_CR61","doi-asserted-by":"publisher","first-page":"596","DOI":"10.1162\/neco_a_01648","volume":"36","author":"VR Anisetti","year":"2024","unstructured":"Anisetti, V. R., Kandala, A., Scellier, B. & Schwarz, J. Frequency propagation: multimechanism learning in nonlinear physical networks. Neural Comput. 36, 596\u2013620 (2024).","journal-title":"Neural Comput."},{"key":"9384_CR62","unstructured":"Murugan, A., Strupp, A., Scellier, B. & Falk, M. Contrastive learning through non-equilibrium memory. In APS March Meeting Abstracts 2023, F02.005 (APS, 2023)."},{"key":"9384_CR63","unstructured":"Laborieux, A. & Zenke, F. Holomorphic equilibrium propagation computes exact gradients through finite size oscillations. In Proc. 36th International Conference on Neural Information Processing Systems (NIPS \u201922), 12950\u201312963 (ACM, 2022)."},{"key":"9384_CR64","unstructured":"Scellier, B., Mishra, S., Bengio, Y. & Ollivier, Y. Agnostic physics-driven deep learning. Preprint at https:\/\/arxiv.org\/abs\/2205.15021 (2022)."},{"key":"9384_CR65","first-page":"031020","volume":"13","author":"V Lopez-Pastor","year":"2023","unstructured":"Lopez-Pastor, V. & Marquardt, F. Self-learning machines based on Hamiltonian echo backpropagation. Phys. Rev. X 13, 031020 (2023).","journal-title":"Phys. Rev. X"},{"key":"9384_CR66","unstructured":"Touvron, H. et al. LLaMA: open and efficient foundation language models. Preprint at https:\/\/arxiv.org\/abs\/2302.13971 (2023)."},{"key":"9384_CR67","first-page":"1","volume":"24","author":"A Chowdhery","year":"2023","unstructured":"Chowdhery, A. et al. PaLM: scaling language modeling with pathways. J. Mach. Learn. Res. 24, 1\u2013113 (2023).","journal-title":"J. Mach. Learn. Res."},{"key":"9384_CR68","unstructured":"Achiam, J. et al. GPT-4 technical report. Preprint at https:\/\/arxiv.org\/abs\/2303.08774v1 (2023)."},{"key":"9384_CR69","unstructured":"Team, G. Gemini: a family of highly capable multimodal models. Preprint at https:\/\/arxiv.org\/abs\/2312.11805v1 (2024)."},{"key":"9384_CR70","unstructured":"Radford, A. et al. Learning transferable visual models from natural language supervision. In Proc. 38th International Conference on Machine Learning, 8748\u20138763 (MLR Press, 2021)."},{"key":"9384_CR71","unstructured":"Liu, H., Li, C., Wu, Q. & Lee, Y. J. Visual instruction tuning. In Proc. 37th Conference on Neural Information Processing Systems (NeurIPS 2023) https:\/\/openreview.net\/forum?id=w0H2xGHlkw (OpenReview, 2023)."},{"key":"9384_CR72","first-page":"9","volume":"1","author":"A Radford","year":"2019","unstructured":"Radford, A. et al. Language models are unsupervised multitask learners. OpenAI Blog 1, 9 (2019).","journal-title":"OpenAI Blog"},{"key":"9384_CR73","unstructured":"Katharopoulos, A., Vyas, A., Pappas, N. & Fleuret, F. Transformers are RNNs: fast autoregressive transformers with linear attention. In Proc. 37th International Conference on Machine Learning, 5156\u20135165 (MLR Press, 2020)."},{"key":"9384_CR74","unstructured":"Gu, A. & Dao, T. Mamba: linear-time sequence modeling with selective state spaces. Preprint at https:\/\/arxiv.org\/abs\/2312.00752v1 (2023)."},{"key":"9384_CR75","unstructured":"Wang, H. et al. BitNet: scaling 1-bit transformers for large language models. Preprint at https:\/\/arxiv.org\/abs\/2310.11453 (2023)."},{"key":"9384_CR76","unstructured":"Hu, E. J. et al. LoRA: low-rank adaptation of large language models. Preprint at https:\/\/arxiv.org\/abs\/2106.09685 (2021)."},{"key":"9384_CR77","unstructured":"Dao, T., Fu, D., Ermon, S., Rudra, A. & R\u00e9, C. FLASHATTENTION: fast and memory-efficient exact attention with IO-awareness. In Proc. 36th Conference on Neural Information Processing Systems (NeurIPS 2022) 35, 16344\u201316359 (ACM, 2022)."},{"key":"9384_CR78","unstructured":"Juravsky, J. et al. Hydragen: high-throughput LLM inference with shared prefixes. Preprint at https:\/\/arxiv.org\/abs\/2402.05099 (2024)."},{"key":"9384_CR79","unstructured":"Anderson, M. G., Ma, S.-Y., Wang, T., Wright, L. G. & McMahon, P. L. Optical transformers. Preprint at https:\/\/arxiv.org\/abs\/2302.10360 (2023)."},{"key":"9384_CR80","doi-asserted-by":"publisher","first-page":"441","DOI":"10.1038\/nphoton.2017.93","volume":"11","author":"Y Shen","year":"2017","unstructured":"Shen, Y. et al. Deep learning with coherent nanophotonic circuits. Nat. Photon. 11, 441\u2013446 (2017).","journal-title":"Nat. Photon."},{"key":"9384_CR81","first-page":"021032","volume":"9","author":"R Hamerly","year":"2019","unstructured":"Hamerly, R., Bernstein, L., Sludds, A., Solja\u010di\u0107, M. & Englund, D. Large-scale optical neural networks based on photoelectric multiplication. Phys. Rev. X 9, 021032 (2019).","journal-title":"Phys. Rev. X"},{"key":"9384_CR82","doi-asserted-by":"publisher","first-page":"054029","DOI":"10.1103\/PhysRevApplied.17.054029","volume":"17","author":"AN Tait","year":"2022","unstructured":"Tait, A. N. Quantifying power in silicon photonic neural networks. Phys. Rev. Appl. 17, 054029 (2022).","journal-title":"Phys. Rev. Appl."},{"key":"9384_CR83","doi-asserted-by":"publisher","first-page":"180","DOI":"10.1016\/j.neuron.2023.11.004","volume":"112","author":"J Laydevant","year":"2024","unstructured":"Laydevant, J., Wright, L. G., Wang, T. & McMahon, P. L. The hardware is the software. Neuron 112, 180\u2013183 (2024).","journal-title":"Neuron"},{"key":"9384_CR84","doi-asserted-by":"publisher","first-page":"58","DOI":"10.1145\/3467017","volume":"64","author":"S Hooker","year":"2021","unstructured":"Hooker, S. The hardware lottery. Commun. ACM 64, 58\u201365 (2021).","journal-title":"Commun. ACM"},{"key":"9384_CR85","doi-asserted-by":"publisher","first-page":"2300055","DOI":"10.1002\/qute.202300055","volume":"6","author":"N Stroev","year":"2023","unstructured":"Stroev, N. & Berloff, N. G. Analog photonics computing for information processing, inference, and optimization. Adv. Quantum Technol. 6, 2300055 (2023).","journal-title":"Adv. Quantum Technol."},{"key":"9384_CR86","doi-asserted-by":"publisher","first-page":"567","DOI":"10.1038\/s43588-022-00311-3","volume":"2","author":"M Cerezo","year":"2022","unstructured":"Cerezo, M., Verdon, G., Huang, H.-Y., Cincio, L. & Coles, P. J. Challenges and opportunities in quantum machine learning. Nat. Comput. Sci. 2, 567\u2013576 (2022).","journal-title":"Nat. Comput. Sci."},{"key":"9384_CR87","doi-asserted-by":"crossref","unstructured":"Kashif, M. & Shafique, M. Hqnet: harnessing quantum noise for effective training of quantum neural networks in NISQ era. Preprint at https:\/\/arxiv.org\/abs\/2402.08475v1 (2024).","DOI":"10.1109\/IJCNN60899.2024.10651363"},{"key":"9384_CR88","doi-asserted-by":"publisher","first-page":"0134","DOI":"10.34133\/research.0134","volume":"6","author":"M-G Zhou","year":"2023","unstructured":"Zhou, M.-G. et al. Quantum neural network for quantum neural computing. Research 6, 0134 (2023).","journal-title":"Research"},{"key":"9384_CR89","doi-asserted-by":"publisher","first-page":"12321","DOI":"10.1109\/TPAMI.2023.3272029","volume":"45","author":"J Tian","year":"2023","unstructured":"Tian, J. et al. Recent advances for quantum neural networks in generative learning. IEEE Trans. Pattern. Anal. Mach. Intell. 45, 12321\u201312340 (2023).","journal-title":"IEEE Trans. Pattern. Anal. Mach. Intell."},{"key":"9384_CR90","doi-asserted-by":"publisher","first-page":"625","DOI":"10.1038\/s42254-021-00348-9","volume":"3","author":"M Cerezo","year":"2021","unstructured":"Cerezo, M. et al. Variational quantum algorithms. Nat. Rev. Phys. 3, 625\u2013644 (2021).","journal-title":"Nat. Rev. Phys."},{"key":"9384_CR91","doi-asserted-by":"publisher","first-page":"610","DOI":"10.1038\/s41928-024-01182-4","volume":"7","author":"S Niazi","year":"2024","unstructured":"Niazi, S. et al. Training deep Boltzmann networks with sparse Ising machines. Nat. Electron. 7, 610\u2013619 (2024).","journal-title":"Nat. Electron."},{"key":"9384_CR92","doi-asserted-by":"crossref","unstructured":"Ma, S. Y., Wang, T., Laydevant, J., Wright, L. G. & McMahon, P. L. Quantum-limited stochastic optical neural networks operating at a few quanta per activation. Nat. Commun. 16, 359 (2025).","DOI":"10.1038\/s41467-024-55220-y"},{"key":"9384_CR93","doi-asserted-by":"publisher","first-page":"4109","DOI":"10.1515\/nanoph-2020-0119","volume":"9","author":"D Pierangeli","year":"2020","unstructured":"Pierangeli, D., Marcucci, G., Brunner, D. & Conti, C. Noise-enhanced spatial-photonic Ising machine. Nanophotonics 9, 4109\u20134116 (2020).","journal-title":"Nanophotonics"},{"key":"9384_CR94","doi-asserted-by":"publisher","first-page":"717","DOI":"10.1038\/s42254-023-00645-5","volume":"5","author":"PL McMahon","year":"2023","unstructured":"McMahon, P. L. The physics of optical computing. Nat. Rev. Phys. 5, 717\u2013734 (2023).","journal-title":"Nat. Rev. Phys."},{"key":"9384_CR95","doi-asserted-by":"publisher","first-page":"131","DOI":"10.1080\/00107514.2010.550120","volume":"52","author":"J Keeling","year":"2011","unstructured":"Keeling, J. & Berloff, N. G. Exciton\u2013polariton condensation. Contemp. Phys. 52, 131\u2013151 (2011).","journal-title":"Contemp. Phys."},{"key":"9384_CR96","doi-asserted-by":"publisher","first-page":"1120","DOI":"10.1038\/nmat4971","volume":"16","author":"NG Berloff","year":"2017","unstructured":"Berloff, N. G. et al. Realizing the classical XY Hamiltonian in polariton simulators. Nat. Mater. 16, 1120\u20131126 (2017).","journal-title":"Nat. Mater."},{"key":"9384_CR97","doi-asserted-by":"publisher","first-page":"096901","DOI":"10.1103\/PhysRevLett.132.096901","volume":"132","author":"A Johnston","year":"2024","unstructured":"Johnston, A. & Berloff, N. G. Macroscopic noise amplification by asymmetric dyads in non-Hermitian optical systems for generative diffusion models. Phys. Rev. Lett. 132, 096901 (2024).","journal-title":"Phys. Rev. Lett."},{"key":"9384_CR98","doi-asserted-by":"publisher","first-page":"408","DOI":"10.1038\/s41566-023-01170-8","volume":"17","author":"T Wang","year":"2023","unstructured":"Wang, T. et al. Image sensing with multilayer nonlinear optical neural networks. Nat. Photon. 17, 408\u2013415 (2023).","journal-title":"Nat. Photon."},{"key":"9384_CR99","doi-asserted-by":"publisher","first-page":"664","DOI":"10.1038\/s41928-020-00501-9","volume":"3","author":"F Zhou","year":"2020","unstructured":"Zhou, F. & Chai, Y. Near-sensor and in-sensor computing. Nat. Electron. 3, 664\u2013671 (2020).","journal-title":"Nat. Electron."},{"key":"9384_CR100","doi-asserted-by":"publisher","first-page":"1901913","DOI":"10.1002\/advs.201901913","volume":"7","author":"P del Hougne","year":"2020","unstructured":"del Hougne, P., F. Imani, M., Diebold, A. V., Horstmeyer, R. & Smith, D. R. Learned integrated sensing pipeline: reconfigurable metasurface transceivers as trainable physical layer in an artificial neural network. Adv. Sci. 7, 1901913 (2020).","journal-title":"Adv. Sci."},{"key":"9384_CR101","unstructured":"Vaswani, A. et al. Attention is all you need. In Proc. 31st International Conference on Neural Information Processing Systems (NIPS \u201917), 6000\u20136010 (ACM, 2017)."},{"key":"9384_CR102","doi-asserted-by":"publisher","first-page":"eabm2956","DOI":"10.1126\/sciadv.abm2956","volume":"8","author":"C Wu","year":"2022","unstructured":"Wu, C. et al. Harnessing optoelectronic noises in a photonic generative network. Sci. Adv. 8, eabm2956 (2022).","journal-title":"Sci. Adv."},{"key":"9384_CR103","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-43317-9","volume":"14","author":"D Bonnet","year":"2023","unstructured":"Bonnet, D. et al. Bringing uncertainty quantification to the extreme-edge with memristor-based Bayesian neural networks. Nat. Commun. 14, 7530 (2023).","journal-title":"Nat. Commun."},{"key":"9384_CR104","doi-asserted-by":"publisher","first-page":"23","DOI":"10.1016\/j.neucom.2020.07.105","volume":"419","author":"W Olin-Ammentorp","year":"2021","unstructured":"Olin-Ammentorp, W., Beckmann, K., Schuman, C. D., Plank, J. S. & Cady, N. C. Stochasticity and robustness in spiking neural networks. Neurocomputing 419, 23\u201336 (2021).","journal-title":"Neurocomputing"}],"container-title":["Nature"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41586-025-09384-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41586-025-09384-2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41586-025-09384-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,9,5]],"date-time":"2025-09-05T11:02:33Z","timestamp":1757070153000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41586-025-09384-2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9,3]]},"references-count":104,"journal-issue":{"issue":"8079","published-print":{"date-parts":[[2025,9,4]]}},"alternative-id":["9384"],"URL":"https:\/\/doi.org\/10.1038\/s41586-025-09384-2","relation":{},"ISSN":["0028-0836","1476-4687"],"issn-type":[{"value":"0028-0836","type":"print"},{"value":"1476-4687","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,9,3]]},"assertion":[{"value":"25 January 2024","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"10 July 2025","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"3 September 2025","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"5 September 2025","order":4,"name":"change_date","label":"Change Date","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"Update","order":5,"name":"change_type","label":"Change Type","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"In the version of this article initially published, affiliation 8 listed outdated information and the affiliation is now updated in the HTML and PDF versions of the article.","order":6,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"T.O., L.G.W. and P.L.M. are listed as inventors on a US provisional patent application (number 63\/178,318) on physical neural networks and physics-aware training.","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}]}}