{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,7,6]],"date-time":"2025-07-06T01:27:37Z","timestamp":1751765257885,"version":"3.37.3"},"reference-count":75,"publisher":"IOP Publishing","issue":"4","license":[{"start":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T00:00:00Z","timestamp":1732147200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T00:00:00Z","timestamp":1732147200000},"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":["Neuromorph. Comput. Eng."],"published-print":{"date-parts":[[2024,12,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Designing algorithms for versatile AI hardware that can learn on the edge using both labeled and unlabeled data is challenging. Deep end-to-end training methods incorporating phases of self-supervised and supervised learning are accurate and adaptable to input data but self-supervised learning requires even more computational and memory resources than supervised learning, too high for current embedded hardware. Conversely, unsupervised layer-by-layer training, such as Hebbian learning, is more compatible with existing hardware but does not integrate well with supervised learning. To address this, we propose a method enabling networks or hardware designed for end-to-end supervised learning to also perform high-performance unsupervised learning by adding two simple elements to the output layer: winner-take-all selectivity and homeostasis regularization. These mechanisms introduce a \u2018self-defined target\u2019 for unlabeled data, allowing purely unsupervised training for both fully-connected and convolutional layers using backpropagation or equilibrium propagation on datasets like MNIST (up to 99.2%), Fashion-MNIST (up to 90.3%), and SVHN (up to 81.5%). We extend this method to semi-supervised learning, adjusting targets based on data type, achieving 96.6% accuracy with only 600 labeled MNIST samples in a multi-layer perceptron. Our results show that this approach can effectively enable networks and hardware initially dedicated to supervised learning to also perform unsupervised learning, adapting to varying availability of labeled data.<\/jats:p>","DOI":"10.1088\/2634-4386\/ad8c78","type":"journal-article","created":{"date-parts":[[2024,10,29]],"date-time":"2024-10-29T22:54:28Z","timestamp":1730242468000},"page":"044005","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Unsupervised end-to-end training with a self-defined target"],"prefix":"10.1088","volume":"4","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6668-4925","authenticated-orcid":true,"given":"Dongshu","family":"Liu","sequence":"first","affiliation":[]},{"given":"J\u00e9r\u00e9mie","family":"Laydevant","sequence":"additional","affiliation":[]},{"given":"Adrien","family":"Pontlevy","sequence":"additional","affiliation":[]},{"given":"Damien","family":"Querlioz","sequence":"additional","affiliation":[]},{"given":"Julie","family":"Grollier","sequence":"additional","affiliation":[]}],"member":"266","published-online":{"date-parts":[[2024,11,21]]},"reference":[{"key":"ncead8c78bib1","doi-asserted-by":"publisher","first-page":"2623","DOI":"10.1145\/3292500.3330701","article-title":"Optuna: a next-generation hyperparameter optimization framework","author":"Akiba","year":"2019"},{"article-title":"Self-labelling via simultaneous clustering and representation learning","year":"2019","author":"Asano","key":"ncead8c78bib2"},{"key":"ncead8c78bib3","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/2658998","article-title":"Programmable spike-timing-dependent plasticity learning circuits in neuromorphic vlsi architectures","volume":"12","author":"Azghadi","year":"2015","journal-title":"ACM J. Emerg. Technol. Comput. Syst."},{"article-title":"A cookbook of self-supervised learning","year":"2023","author":"Balestriero","key":"ncead8c78bib4"},{"key":"ncead8c78bib5","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2105.04906","article-title":"VICREG: variance-invariance-covariance regularization for self-supervised learning","author":"Bardes","year":"2022"},{"key":"ncead8c78bib6","doi-asserted-by":"publisher","first-page":"555","DOI":"10.1109\/iThings-GreenCom-CPSCom-SmartData-Cybermatics55523.2022.00108","article-title":"Semi-supervised algorithms in resource-constrained edge devices: an overview and experimental comparison","author":"Barhoush","year":"2022"},{"key":"ncead8c78bib7","doi-asserted-by":"publisher","DOI":"10.7551\/mitpress\/7503.003.0024","article-title":"Greedy layer-wise training of deep networks","volume":"vol 19","author":"Bengio","year":"2006"},{"key":"ncead8c78bib8","doi-asserted-by":"publisher","first-page":"10464","DOI":"10.1523\/JNEUROSCI.18-24-10464.1998","article-title":"Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength and postsynaptic cell type","volume":"18","author":"Bi","year":"1998","journal-title":"J. Neurosci."},{"key":"ncead8c78bib9","doi-asserted-by":"publisher","first-page":"339","DOI":"10.1016\/j.neunet.2012.02.022","article-title":"Extraction of temporally correlated features from dynamic vision sensors with spike-timing-dependent plasticity","volume":"32","author":"Bichler","year":"2012","journal-title":"Neural Netw."},{"article-title":"A pseudo-labelling auto-encoder for unsupervised image classification","year":"2020","author":"Bouayed","key":"ncead8c78bib10"},{"key":"ncead8c78bib11","doi-asserted-by":"publisher","first-page":"85714","DOI":"10.1109\/ACCESS.2020.2991734","article-title":"An overview on edge computing research","volume":"8","author":"Cao","year":"2020","journal-title":"IEEE Access"},{"article-title":"Curriculum labeling: revisiting pseudo-labeling for semi-supervised learning","year":"2020","author":"Cascante-Bonilla","key":"ncead8c78bib12"},{"key":"ncead8c78bib13","doi-asserted-by":"publisher","first-page":"440","DOI":"10.1109\/sibgrapi.2018.00063","article-title":"Unsupervised representation learning using convolutional and stacked auto-encoders: a domain and cross-domain feature space analysis","author":"Cavallari","year":"2018"},{"key":"ncead8c78bib14","doi-asserted-by":"publisher","first-page":"1655","DOI":"10.1109\/JPROC.2019.2921977","article-title":"Deep learning with edge computing: a review","volume":"107","author":"Chen","year":"2019","journal-title":"Proc. IEEE"},{"key":"ncead8c78bib15","first-page":"215","article-title":"An analysis of single-layer networks in unsupervised feature learning","author":"Coates","year":"2011"},{"key":"ncead8c78bib16","doi-asserted-by":"publisher","first-page":"99","DOI":"10.3389\/fncom.2015.00099","article-title":"Unsupervised learning of digit recognition using spike-timing-dependent plasticity","volume":"9","author":"Diehl","year":"2015","journal-title":"Front. Comput. Neurosci."},{"key":"ncead8c78bib17","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevApplied.18.014040","article-title":"Demonstration of decentralized physics-driven learning","volume":"18","author":"Dillavou","year":"2022","journal-title":"Phys. Rev. Appl."},{"year":"1949","author":"Do","key":"ncead8c78bib18"},{"key":"ncead8c78bib19","first-page":"3","article-title":"Visualizing higher-layer features of a deep network","volume":"1341","author":"Erhan","year":"2009"},{"key":"ncead8c78bib20","doi-asserted-by":"publisher","first-page":"49393","DOI":"10.1109\/ACCESS.2023.3277253","article-title":"A novel similarity based unsupervised technique for training convolutional filters","volume":"11","author":"Erko\u00e7","year":"2023","journal-title":"IEEE Access"},{"key":"ncead8c78bib21","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1905.13633","article-title":"Updates of equilibrium prop match gradients of backprop through time in an RNN with static input","volume":"vol 32","author":"Ernoult","year":"2019"},{"article-title":"Equilibrium propagation with continual weight updates","year":"2020","author":"Ernoult","key":"ncead8c78bib22"},{"key":"ncead8c78bib23","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3389\/fncom.2018.00024","article-title":"Unsupervised feature learning with winner-takes-all based stdp","volume":"12","author":"Ferr\u00e9","year":"2018","journal-title":"Front. Comput. Neurosci."},{"key":"ncead8c78bib24","doi-asserted-by":"publisher","DOI":"10.7566\/JPSJ.90.074802","article-title":"Semi-supervised learning combining backpropagation and STDP: STDP enhances learning by backpropagation with a small amount of labeled data in a spiking neural network","volume":"90","author":"Furuya","year":"2021","journal-title":"J. Phys. Soc. Japan"},{"article-title":"Is bio-inspired learning better than backprop? Benchmarking bio learning vs. backprop","year":"2022","author":"Gupta","key":"ncead8c78bib25"},{"key":"ncead8c78bib26","doi-asserted-by":"publisher","first-page":"770","DOI":"10.1109\/CVPR.2016.90","article-title":"Deep residual learning for image recognition","author":"He","year":"2016"},{"article-title":"Improving neural networks by preventing co-adaptation of feature detectors","year":"2012","author":"Hinton","key":"ncead8c78bib27"},{"key":"ncead8c78bib28","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1109\/TNN.2005.860850","article-title":"A vlsi array of low-power spiking neurons and bistable synapses with spike-timing dependent plasticity","volume":"17","author":"Indiveri","year":"2006","journal-title":"IEEE Trans. Neural Netw."},{"article-title":"Batch normalization: accelerating deep network training by reducing internal covariate shift","year":"2015","author":"Ioffe","key":"ncead8c78bib29"},{"key":"ncead8c78bib30","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1109\/IEDM19573.2019.8993466","article-title":"On-chip trainable 1.4 m 6T2R PCM synaptic array with 1.6 K stochastic LIF neurons for spiking RBM","author":"Ishii","year":"2019"},{"key":"ncead8c78bib31","doi-asserted-by":"publisher","first-page":"1297","DOI":"10.1021\/nl904092h","article-title":"Nanoscale memristor device as synapse in neuromorphic systems","volume":"10","author":"Jo","year":"2010","journal-title":"Nano Lett."},{"article-title":"Hebbian deep learning without feedback","year":"2023","author":"Journ\u00e9","key":"ncead8c78bib32"},{"article-title":"Training end-to-end analog neural networks with equilibrium propagation","year":"2020","author":"Kendall","key":"ncead8c78bib33"},{"article-title":"Learning multiple layers of features from tiny images","year":"2009","author":"Krizhevsky","key":"ncead8c78bib34"},{"key":"ncead8c78bib35","doi-asserted-by":"publisher","first-page":"7723","DOI":"10.1073\/pnas.1820458116","article-title":"Unsupervised learning by competing hidden units","volume":"116","author":"Krotov","year":"2019","journal-title":"Proc. Natl Acad. Sci. USA"},{"key":"ncead8c78bib36","doi-asserted-by":"publisher","DOI":"10.3389\/fnins.2021.633674","article-title":"Scaling equilibrium propagation to deep convnets by drastically reducing its gradient estimator bias","volume":"15","author":"Laborieux","year":"2021","journal-title":"Front. Neurosci."},{"key":"ncead8c78bib37","doi-asserted-by":"publisher","first-page":"12950","DOI":"10.48550\/arXiv.2209.00530","article-title":"Holomorphic equilibrium propagation computes exact gradients through finite size oscillations","volume":"vol 35","author":"Laborieux","year":"2022"},{"key":"ncead8c78bib38","doi-asserted-by":"publisher","first-page":"365","DOI":"10.1007\/978-3-030-95470-3_28","article-title":"Evaluating Hebbian learning in a semi-supervised setting","author":"Lagani","year":"2021"},{"key":"ncead8c78bib39","doi-asserted-by":"publisher","first-page":"6503","DOI":"10.1007\/s00521-021-06701-4","article-title":"Comparing the performance of Hebbian against backpropagation learning using convolutional neural networks","volume":"34","author":"Lagani","year":"2022","journal-title":"Neural Comput. Appl."},{"article-title":"Synaptic plasticity models and bio-inspired unsupervised deep learning: a survey","year":"2023","author":"Lagani","key":"ncead8c78bib40"},{"key":"ncead8c78bib41","doi-asserted-by":"publisher","first-page":"1","DOI":"10.48550\/arXiv.1610.02242","article-title":"Temporal ensembling for semi-supervised learning","author":"Laine","year":"2016"},{"article-title":"Training an ising machine with equilibrium propagation","year":"2023","author":"Laydevant","key":"ncead8c78bib42"},{"article-title":"MNIST handwritten digit database","year":"2010","author":"LeCun","key":"ncead8c78bib43"},{"article-title":"Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks","year":"2013","author":"Lee","key":"ncead8c78bib44"},{"key":"ncead8c78bib45","doi-asserted-by":"publisher","first-page":"35140","DOI":"10.1109\/ACCESS.2023.3264435","article-title":"Semi-supervised learning for spiking neural networks based on spike-timing-dependent plasticity","volume":"11","author":"Lee","year":"2023","journal-title":"IEEE Access"},{"key":"ncead8c78bib46","doi-asserted-by":"publisher","first-page":"2519","DOI":"10.1162\/089976600300014827","article-title":"On the computational power of winner-take-all","volume":"12","author":"Maass","year":"2000","journal-title":"Neural Comput."},{"key":"ncead8c78bib47","first-page":"634","article-title":"On the k-winners-take-all network","volume":"vol 1","author":"Majani","year":"1988"},{"key":"ncead8c78bib48","doi-asserted-by":"publisher","DOI":"10.1016\/j.isci.2021.102222","article-title":"Eqspike: spike-driven equilibrium propagation for neuromorphic implementations","volume":"24","author":"Martin","year":"2021","journal-title":"iScience"},{"article-title":"Umap: uniform manifold approximation and projection for dimension reduction","year":"2018","author":"McInnes","key":"ncead8c78bib49"},{"article-title":"Hebbian learning with gradients: Hebbian convolutional neural networks with modern deep learning frameworks","year":"2021","author":"Miconi","key":"ncead8c78bib50"},{"key":"ncead8c78bib51","doi-asserted-by":"publisher","DOI":"10.1088\/2634-4386\/aca710","article-title":"SoftHebb: Bayesian inference in unsupervised hebbian soft winner-take-all networks","volume":"2","author":"Moraitis","year":"2022","journal-title":"Neuromorph. Comput. Eng."},{"key":"ncead8c78bib52","doi-asserted-by":"publisher","first-page":"613","DOI":"10.1109\/TNNLS.2018.2826721","article-title":"Bio-inspired digit recognition using reward-modulated stdp","volume":"30","author":"Mozafari","year":"2018","journal-title":"IEEE Trans. Neural Netw. Learn. Syst."},{"key":"ncead8c78bib53","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pcbi.1003037","article-title":"Bayesian computation emerges in generic cortical microcircuits through spike-timing-dependent plasticity","volume":"9","author":"Nessler","year":"2013","journal-title":"PLoS Comput. Biol."},{"article-title":"Reading digits in natural images with unsupervised feature learning","year":"2011","author":"Netzer","key":"ncead8c78bib54"},{"key":"ncead8c78bib55","doi-asserted-by":"publisher","first-page":"1929","DOI":"10.5555\/2627435.2670313","article-title":"Dropout: a simple way to prevent neural networks from overfitting","volume":"15","author":"Nitish","year":"2014","journal-title":"J. Mach. Learn. Res."},{"key":"ncead8c78bib56","doi-asserted-by":"publisher","first-page":"267","DOI":"10.1007\/BF00275687","article-title":"Simplified neuron model as a principal component analyzer","volume":"15","author":"Oja","year":"1982","journal-title":"J. Math. Biol."},{"key":"ncead8c78bib57","doi-asserted-by":"publisher","first-page":"2437","DOI":"10.1162\/neco.2009.07-08-829","article-title":"Computation with spikes in a winner-take-all network","volume":"21","author":"Oster","year":"2009","journal-title":"Neural Comput."},{"key":"ncead8c78bib58","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1701.06548","article-title":"Regularizing neural networks by penalizing confident output distributions","author":"Pereyra","year":"2017"},{"key":"ncead8c78bib59","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/IJCNN48605.2020.9207431","article-title":"Improving image autoencoder embeddings with perceptual loss","author":"Pihlgren","year":"2020"},{"key":"ncead8c78bib60","doi-asserted-by":"publisher","first-page":"288","DOI":"10.1109\/TNANO.2013.2250995","article-title":"Immunity to device variations in a spiking neural network with memristive nanodevices","volume":"12","author":"Querlioz","year":"2013","journal-title":"IEEE Trans. Nanotechnol."},{"key":"ncead8c78bib61","doi-asserted-by":"publisher","first-page":"24","DOI":"10.3389\/fncom.2017.00024","article-title":"Equilibrium propagation: bridging the gap between energy-based models and backpropagation","volume":"11","author":"Scellier","year":"2017","journal-title":"Front. Comput. Neurosci."},{"key":"ncead8c78bib62","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2312.15103","article-title":"Energy-based learning algorithms for analog computing: a comparative study","volume":"vol 36","author":"Scellier","year":"2024"},{"key":"ncead8c78bib63","doi-asserted-by":"publisher","DOI":"10.1016\/j.comnet.2020.107496","article-title":"A survey on the computation offloading approaches in mobile edge computing: a machine learning-based perspective","volume":"182","author":"Shakarami","year":"2020","journal-title":"Comput. Netw."},{"article-title":"Blockwise self-supervised learning at scale","year":"2022","author":"Siddiqui","key":"ncead8c78bib64"},{"key":"ncead8c78bib65","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1703.01780","article-title":"Mean teachers are better role models: weight-averaged consistency targets improve semi-supervised deep learning results","volume":"vol 30","author":"Tarvainen","year":"2017"},{"key":"ncead8c78bib66","doi-asserted-by":"publisher","first-page":"542","DOI":"10.1109\/TNN.2009.2015974","article-title":"Semi-supervised learning (chapelle, o. et al., eds.; 2006)[book reviews]","volume":"20","author":"Thomas","year":"2009","journal-title":"IEEE Trans. Neural Netw."},{"key":"ncead8c78bib67","doi-asserted-by":"publisher","first-page":"197","DOI":"10.1007\/978-3-030-20205-7_17","article-title":"Unsupervised feature extraction\u2013a CNN-based approach","volume":"vol 21","author":"Trosten","year":"2019"},{"key":"ncead8c78bib68","doi-asserted-by":"publisher","first-page":"373","DOI":"10.1007\/s10994-019-05855-6","article-title":"A survey on semi-supervised learning","volume":"109","author":"Van Engelen","year":"2020","journal-title":"Mach. Learn."},{"article-title":"Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms","year":"2017","author":"Xiao","key":"ncead8c78bib69"},{"key":"ncead8c78bib70","doi-asserted-by":"publisher","DOI":"10.1088\/2634-4386\/ac734a","article-title":"Memristive devices based hardware for unlabeled data processing","volume":"2","author":"Xiao","year":"2022","journal-title":"Neuromorph. Comput. Eng."},{"key":"ncead8c78bib71","first-page":"189","article-title":"Unsupervised word sense disambiguation rivaling supervised methods","author":"Yarowsky","year":"1995"},{"key":"ncead8c78bib72","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1038\/s41928-022-00869-w","article-title":"Activity-difference training of deep neural networks using memristor crossbars","volume":"6","author":"Yi","year":"2022","journal-title":"Nat. Electron."},{"key":"ncead8c78bib73","doi-asserted-by":"publisher","first-page":"12310","DOI":"10.48550\/arXiv.2103.03230","article-title":"Barlow twins: self-supervised learning via redundancy reduction","author":"Zbontar","year":"2021"},{"article-title":"Activation learning by local competitions","year":"2022","author":"Zhou","key":"ncead8c78bib74"},{"year":"2022","author":"Zhu","key":"ncead8c78bib75","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-01548-9"}],"container-title":["Neuromorphic Computing and Engineering"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T12:27:49Z","timestamp":1732192069000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2634-4386\/ad8c78"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,11,21]]},"references-count":75,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2024,11,21]]},"published-print":{"date-parts":[[2024,12,1]]}},"URL":"https:\/\/doi.org\/10.1088\/2634-4386\/ad8c78","relation":{},"ISSN":["2634-4386"],"issn-type":[{"type":"electronic","value":"2634-4386"}],"subject":[],"published":{"date-parts":[[2024,11,21]]},"assertion":[{"value":"Unsupervised end-to-end training with a self-defined target","name":"article_title","label":"Article Title"},{"value":"Neuromorphic Computing and Engineering","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2024 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2024-03-18","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2024-10-29","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2024-11-21","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}