{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,26]],"date-time":"2026-03-26T12:20:15Z","timestamp":1774527615914,"version":"3.50.1"},"reference-count":73,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2023,5,24]],"date-time":"2023-05-24T00:00:00Z","timestamp":1684886400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100001711","name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001711","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100004412","name":"Human Frontier Science Program","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100004412","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100003006","name":"Eidgen\u00f6ssische Technische Hochschule Z\u00fcrich","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100003006","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Comput. Neurosci."],"abstract":"<jats:p>A key driver of mammalian intelligence is the ability to represent incoming sensory information across multiple abstraction levels. For example, in the visual ventral stream, incoming signals are first represented as low-level edge filters and then transformed into high-level object representations. Similar hierarchical structures routinely emerge in artificial neural networks (ANNs) trained for object recognition tasks, suggesting that similar structures may underlie biological neural networks. However, the classical ANN training algorithm, backpropagation, is considered biologically implausible, and thus alternative biologically plausible training methods have been developed such as Equilibrium Propagation, Deep Feedback Control, Supervised Predictive Coding, and Dendritic Error Backpropagation. Several of those models propose that local errors are calculated for each neuron by comparing apical and somatic activities. Notwithstanding, from a neuroscience perspective, it is not clear how a neuron could compare compartmental signals. Here, we propose a solution to this problem in that we let the apical feedback signal change the postsynaptic firing rate and combine this with a differential Hebbian update, a rate-based version of classical spiking time-dependent plasticity (STDP). We prove that weight updates of this form minimize two alternative loss functions that we prove to be equivalent to the error-based losses used in machine learning: the inference latency and the amount of top-down feedback necessary. Moreover, we show that the use of differential Hebbian updates works similarly well in other feedback-based deep learning frameworks such as Predictive Coding or Equilibrium Propagation. Finally, our work removes a key requirement of biologically plausible models for deep learning and proposes a learning mechanism that would explain how temporal Hebbian learning rules can implement supervised hierarchical learning.<\/jats:p>","DOI":"10.3389\/fncom.2023.1136010","type":"journal-article","created":{"date-parts":[[2023,5,24]],"date-time":"2023-05-24T06:09:21Z","timestamp":1684908561000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":8,"title":["Learning cortical hierarchies with temporal Hebbian updates"],"prefix":"10.3389","volume":"17","author":[{"given":"Pau Vilimelis","family":"Aceituno","sequence":"first","affiliation":[]},{"given":"Matilde Tristany","family":"Farinha","sequence":"additional","affiliation":[]},{"given":"Reinhard","family":"Loidl","sequence":"additional","affiliation":[]},{"given":"Benjamin F.","family":"Grewe","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2023,5,24]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"1178","DOI":"10.1038\/81453","article-title":"Synaptic plasticity: taming the beast","volume":"3","author":"Abbott","year":"2000","journal-title":"Nat. Neurosci"},{"key":"B2","doi-asserted-by":"publisher","first-page":"202","DOI":"10.1016\/j.neuron.2019.01.036","article-title":"Somatostatin-expressing interneurons enable and maintain learning-dependent sequential activation of pyramidal neurons","volume":"102","author":"Avital","year":"2019","journal-title":"Neuron"},{"key":"B3","doi-asserted-by":"publisher","first-page":"695","DOI":"10.1016\/j.neuron.2012.10.038","article-title":"Canonical microcircuits for predictive coding","volume":"76","author":"Bastos","year":"2012","journal-title":"Neuron"},{"key":"B4","article-title":"Towards biologically plausible deep learning","author":"Bengio","year":"2015","journal-title":"arXiv preprint arXiv:1502.04156"},{"key":"B5","doi-asserted-by":"publisher","first-page":"555","DOI":"10.1162\/NECO_a_00934","article-title":"Stdp-compatible approximation of back-propagation in an energy-based model","volume":"29","author":"Bengio","year":"2017","journal-title":"Neural Comput"},{"key":"B6","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":"B7","article-title":"Precise deep neural network computation on imprecise low-power analog hardware","author":"Binas","year":"2016","journal-title":"Comput. Sci. Neural Evol. Comput"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.1101\/2020.11.02.364968","article-title":"Learning to live with dale's principle: ANNs with separate excitatory and inhibitory units","author":"Cornford","year":"2020","journal-title":"bioRxiv [Preprint]"},{"key":"B9","doi-asserted-by":"publisher","first-page":"129","DOI":"10.1038\/337129a0","article-title":"The recent excitement about neural networks","volume":"337","author":"Crick","year":"1989","journal-title":"Nature"},{"key":"B10","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":"B11","doi-asserted-by":"publisher","first-page":"556","DOI":"10.1016\/j.neuron.2012.08.001","article-title":"The spike-timing dependence of plasticity","volume":"75","author":"Feldman","year":"2012","journal-title":"Neuron"},{"key":"B12","doi-asserted-by":"publisher","first-page":"85","DOI":"10.3389\/fncir.2015.00085","article-title":"Neuromodulated spike-timing-dependent plasticity, and theory of three-factor learning rules","volume":"9","author":"Fr\u00e9maux","year":"2016","journal-title":"Front. Neural Circuits"},{"key":"B13","doi-asserted-by":"publisher","first-page":"1211","DOI":"10.1098\/rstb.2008.0300","article-title":"Predictive coding under the free-energy principle","volume":"364","author":"Friston","year":"2009","journal-title":"Philos. Trans. R. Soc. B Biol. Sci"},{"key":"B14","doi-asserted-by":"publisher","first-page":"453","DOI":"10.1016\/j.clinph.2008.11.029","article-title":"The mismatch negativity: a review of underlying mechanisms","volume":"120","author":"Garrido","year":"2009","journal-title":"Clin. Neurophysiol"},{"key":"B15","doi-asserted-by":"publisher","first-page":"76","DOI":"10.1038\/383076a0","article-title":"A neuronal learning rule for sub-millisecond temporal coding","volume":"383","author":"Gerstner","year":"1996","journal-title":"Nature"},{"key":"B16","doi-asserted-by":"publisher","first-page":"350","DOI":"10.1038\/nrn3476","article-title":"Top-down influences on visual processing","volume":"14","author":"Gilbert","year":"2013","journal-title":"Nat. Rev. Neurosci"},{"key":"B17","doi-asserted-by":"publisher","first-page":"e28295","DOI":"10.7554\/eLife.28295","article-title":"Predicting non-linear dynamics by stable local learning in a recurrent spiking neural network","volume":"6","author":"Gilra","year":"2017","journal-title":"Elife"},{"key":"B18","doi-asserted-by":"publisher","first-page":"aab4113","DOI":"10.1126\/science.aab4113","article-title":"Spiking neurons can discover predictive features by aggregate-label learning","volume":"351","author":"G\u00fctig","year":"2016","journal-title":"Science"},{"key":"B19","doi-asserted-by":"publisher","first-page":"2229","DOI":"10.1016\/j.celrep.2019.04.075","article-title":"Two-color volumetric imaging of neuronal activity of cortical columns","volume":"27","author":"Han","year":"2019","journal-title":"Cell Rep"},{"key":"B20","first-page":"1","article-title":"\u201cHow to do backpropagation in a brain,\u201d","volume-title":"Invited Talk at the NIPS'2007 Deep Learning Workshop, Vol. 656","author":"Hinton","year":"2007"},{"key":"B21","article-title":"Towards truly local gradients with CLAPP: contrastive, local and predictive plasticity","author":"Illing","year":"2020","journal-title":"CoRR, abs\/2010.08262"},{"key":"B22","doi-asserted-by":"publisher","first-page":"90","DOI":"10.1016\/j.neunet.2019.06.001","article-title":"Biologically plausible deep learning-but how far can we go with shallow networks?","volume":"118","author":"Illing","year":"2019","journal-title":"Neural Netw"},{"key":"B23","doi-asserted-by":"publisher","first-page":"33639","DOI":"10.1073\/pnas.2013663117","article-title":"Synaptic plasticity rules with physiological calcium levels","volume":"117","author":"Inglebert","year":"2020","journal-title":"Proc. Natl. Acad. Sci. U.S.A"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1511","DOI":"10.1162\/089976603321891783","article-title":"Relating STDP to BCM","volume":"15","author":"Izhikevich","year":"2003","journal-title":"Neural Comput"},{"key":"B25","doi-asserted-by":"publisher","first-page":"424","DOI":"10.1016\/j.neuron.2018.10.003","article-title":"Predictive processing: a canonical cortical computation","volume":"100","author":"Keller","year":"2018","journal-title":"Neuron"},{"key":"B26","doi-asserted-by":"publisher","first-page":"4498","DOI":"10.1103\/PhysRevE.59.4498","article-title":"Hebbian learning and spiking neurons","volume":"59","author":"Kempter","year":"1999","journal-title":"Phys. Rev. E"},{"key":"B27","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1038\/4511","article-title":"Predicting the visual world: silence is golden","volume":"2","author":"Koch","year":"1999","journal-title":"Nat. Neurosci"},{"key":"B28","doi-asserted-by":"publisher","first-page":"111","DOI":"10.3389\/fncom.2015.00111","article-title":"Is predictive coding theory articulated enough to be testable?","volume":"9","author":"Kogo","year":"2015","journal-title":"Front. Comput. Neurosci"},{"key":"B29","doi-asserted-by":"publisher","first-page":"R231","DOI":"10.1016\/j.cub.2019.02.034","article-title":"Neural network models and deep learning","volume":"29","author":"Kriegeskorte","year":"2019","journal-title":"Curr. Biol"},{"key":"B30","doi-asserted-by":"publisher","first-page":"1262","DOI":"10.1126\/science.1137450","article-title":"Anti-hebbian long-term potentiation in the hippocampal feedback inhibitory circuit","volume":"315","author":"Lamsa","year":"2007","journal-title":"Science"},{"key":"B31","doi-asserted-by":"publisher","first-page":"141","DOI":"10.1016\/j.tins.2012.11.006","article-title":"A cellular mechanism for cortical associations: an organizing principle for the cerebral cortex","volume":"36","author":"Larkum","year":"2013","journal-title":"Trends Neurosci"},{"key":"B32","doi-asserted-by":"publisher","first-page":"1059","DOI":"10.1093\/cercor\/bhh065","article-title":"Top-down dendritic input increases the gain of layer 5 pyramidal neurons","volume":"14","author":"Larkum","year":"2004","journal-title":"Cereb. Cortex"},{"key":"B33","doi-asserted-by":"publisher","first-page":"8999","DOI":"10.1523\/JNEUROSCI.1717-07.2007","article-title":"Dendritic spikes in apical dendrites of neocortical layer 2\/3 pyramidal neurons","volume":"27","author":"Larkum","year":"2007","journal-title":"J. Neurosci"},{"key":"B34","doi-asserted-by":"publisher","first-page":"338","DOI":"10.1038\/18686","article-title":"A new cellular mechanism for coupling inputs arriving at different cortical layers","volume":"398","author":"Larkum","year":"1999","journal-title":"Nature"},{"key":"B35","doi-asserted-by":"publisher","first-page":"23","DOI":"10.3389\/neuro.10.023.2009","article-title":"Sorn: a self-organizing recurrent neural network","volume":"3","author":"Lazar","year":"2009","journal-title":"Front. Comput. Neurosci"},{"key":"B36","unstructured":"LeCunY.\n          The mnist Database of Handwritten Digits1998"},{"key":"B37","doi-asserted-by":"publisher","first-page":"335","DOI":"10.1038\/s41583-020-0277-3","article-title":"Backpropagation and the brain","volume":"21","author":"Lillicrap","year":"2020","journal-title":"Nat. Rev. Neurosci"},{"key":"B38","doi-asserted-by":"publisher","first-page":"210","DOI":"10.1038\/s42256-020-0170-9","article-title":"A neural network trained for prediction mimics diverse features of biological neurons and perception","volume":"2","author":"Lotter","year":"2020","journal-title":"Nat. Mach. Intell"},{"key":"B39","doi-asserted-by":"publisher","first-page":"62","DOI":"10.1038\/s42256-021-00430-y","article-title":"Neurons learn by predicting future activity","volume":"4","author":"Luczak","year":"2022","journal-title":"Nat. Mach. Intell"},{"key":"B40","doi-asserted-by":"publisher","first-page":"213","DOI":"10.1126\/science.275.5297.213","article-title":"Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs","volume":"275","author":"Markram","year":"1997","journal-title":"Science"},{"key":"B41","doi-asserted-by":"publisher","first-page":"1259","DOI":"10.1162\/neco.2008.06-08-804","article-title":"Competitive STDP-based spike pattern learning","volume":"21","author":"Masquelier","year":"2009","journal-title":"Neural Comput"},{"key":"B42","first-page":"20024","article-title":"A theoretical framework for target propagation","volume":"33","author":"Meulemans","year":"2020","journal-title":"Adv. Neural Inform. Process. Syst"},{"key":"B43","first-page":"15458","article-title":"\u201cMinimizing control for credit assignment with strong feedback,\u201d","volume-title":"International Conference on Machine Learning","author":"Meulemans","year":""},{"key":"B44","article-title":"Credit assignment in neural networks through deep feedback control","author":"Meulemans","year":"","journal-title":"arXiv preprint arXiv:2106.07887"},{"key":"B45","article-title":"Credit assignment in neural networks through deep feedback control","author":"Meulemans","year":"","journal-title":"CoRR, abs\/2106.07887"},{"key":"B46","first-page":"33603","article-title":"The least-control principle for local learning at equilibrium","volume":"35","author":"Meulemans","year":"","journal-title":"Adv. Neural Infm. Process. Syst"},{"key":"B47","doi-asserted-by":"publisher","first-page":"625","DOI":"10.3389\/fnins.2019.00625","article-title":"Spyketorch: efficient simulation of convolutional spiking neural networks with at most one spike per neuron","volume":"13","author":"Mozafari","year":"2019","journal-title":"Front. Neurosci"},{"key":"B48","doi-asserted-by":"publisher","first-page":"584","DOI":"10.1038\/35046067","article-title":"Calcium stores regulate the polarity and input specificity of synaptic modification","volume":"408","author":"Nishiyama","year":"2000","journal-title":"Nature"},{"key":"B49","doi-asserted-by":"publisher","first-page":"1546","DOI":"10.1038\/s41593-021-00857-x","article-title":"Burst-dependent synaptic plasticity can coordinate learning in hierarchical circuits","volume":"24","author":"Payeur","year":"2021","journal-title":"Nat. Neurosci"},{"key":"B50","doi-asserted-by":"publisher","first-page":"1433","DOI":"10.1109\/JSSC.1989.572629","article-title":"Matching properties of mos transistors","volume":"24","author":"Pelgrom","year":"1989","journal-title":"IEEE J. Solid-State Circuits"},{"key":"B51","doi-asserted-by":"publisher","first-page":"79","DOI":"10.1038\/4580","article-title":"Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects","volume":"2","author":"Rao","year":"1999","journal-title":"Nat. Neurosci"},{"key":"B52","doi-asserted-by":"publisher","first-page":"e0266102","DOI":"10.1371\/journal.pone.0266102","article-title":"On the relationship between predictive coding and backpropagation","volume":"17","author":"Rosenbaum","year":"2022","journal-title":"PLoS ONE"},{"key":"B53","first-page":"8721","article-title":"\u201cDendritic cortical microcircuits approximate the backpropagation algorithm,\u201d","author":"Sacramento","year":"2018","journal-title":"Advances in Neural Information Processing Systems, Vol. 31"},{"key":"B54","doi-asserted-by":"publisher","DOI":"10.1101\/2021.10.31.466667","article-title":"Sequence anticipation and STDP emerge from a voltage-based predictive learning rule","author":"Saponati","year":"2021","journal-title":"bioRxiv [Preprint]"},{"key":"B55","doi-asserted-by":"publisher","first-page":"595","DOI":"10.1162\/089976604772744929","article-title":"How the shape of pre-and postsynaptic signals can influence STDP: a biophysical model","volume":"16","author":"Saudargiene","year":"2004","journal-title":"Neural Comput"},{"key":"B56","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":"B57","doi-asserted-by":"publisher","first-page":"2694","DOI":"10.1152\/jn.00447.2018","article-title":"Calcium stores regulate excitability in cultured rat hippocampal neurons","volume":"120","author":"Segal","year":"2018","journal-title":"J. Neurophysiol"},{"key":"B58","doi-asserted-by":"publisher","first-page":"1362","DOI":"10.4249\/scholarpedia.1362","article-title":"Spike-timing dependent plasticity","volume":"5","author":"Sj\u00f6str\u00f6m","year":"2010","journal-title":"Scholarpedia"},{"key":"B59","doi-asserted-by":"publisher","first-page":"1149","DOI":"10.1016\/S0896-6273(01)00542-6","article-title":"Rate, timing, and cooperativity jointly determine cortical synaptic plasticity","volume":"32","author":"Sj\u00f6str\u00f6m","year":"2001","journal-title":"Neuron"},{"key":"B60","doi-asserted-by":"publisher","first-page":"919","DOI":"10.1038\/78829","article-title":"Competitive hebbian learning through spike-timing-dependent synaptic plasticity","volume":"3","author":"Song","year":"2000","journal-title":"Nat. Neurosci"},{"key":"B61","doi-asserted-by":"publisher","first-page":"5239","DOI":"10.1073\/pnas.0500495102","article-title":"Generalized bienenstock-cooper-munro rule for spiking neurons that maximizes information transmission","volume":"102","author":"Toyoizumi","year":"2005","journal-title":"Proc. Natl. Acad. Sci. U.S.A"},{"key":"B62","unstructured":"TschantzA.\n          A Python Implementation of An Approximation of the Error Backpropagation Algorithm in a Predictive Coding Network with Local Hebbian Synaptic Plasticity2020"},{"key":"B63","doi-asserted-by":"publisher","first-page":"8812","DOI":"10.1523\/JNEUROSCI.20-23-08812.2000","article-title":"Stable hebbian learning from spike timing-dependent plasticity","volume":"20","author":"van Rossum","year":"2000","journal-title":"J. Neurosci"},{"key":"B64","doi-asserted-by":"publisher","first-page":"43","DOI":"10.1007\/s00422-019-00813-w","article-title":"Spiking time-dependent plasticity leads to efficient coding of predictions","volume":"114","author":"Vilimelis Aceituno","year":"2020","journal-title":"Biol. Cybern"},{"key":"B65","doi-asserted-by":"publisher","first-page":"1229","DOI":"10.1162\/NECO_a_00949","article-title":"An approximation of the error backpropagation algorithm in a predictive coding network with local hebbian synaptic plasticity","volume":"29","author":"Whittington","year":"2017","journal-title":"Neural Comput"},{"key":"B66","doi-asserted-by":"publisher","first-page":"91","DOI":"10.1016\/j.neuron.2018.10.049","article-title":"Higher-order thalamocortical inputs gate synaptic long-term potentiation via disinhibition","volume":"101","author":"Williams","year":"2019","journal-title":"Neuron"},{"key":"B67","doi-asserted-by":"publisher","first-page":"6610","DOI":"10.1523\/JNEUROSCI.5388-05.2006","article-title":"Malleability of spike-timing-dependent plasticity at the CA3-CA1 synapse","volume":"26","author":"Wittenberg","year":"2006","journal-title":"J. Neurosci"},{"key":"B68","article-title":"\u201cSpike-based learning rules and stabilization of persistent neural activity,\u201d","volume-title":"Advances in Neural Information Processing Systems, Vol. 12","author":"Xie","year":"1999"},{"key":"B69","doi-asserted-by":"publisher","first-page":"850945","DOI":"10.3389\/fnins.2022.850945","article-title":"Sam: a unified self-adaptive multicompartmental spiking neuron model for learning with working memory","volume":"16","author":"Yang","year":"","journal-title":"Front. Neurosci"},{"key":"B70","doi-asserted-by":"publisher","first-page":"850932","DOI":"10.3389\/fnins.2022.850932","article-title":"Heterogeneous ensemble-based spike-driven few-shot online learning","volume":"16","author":"Yang","year":"","journal-title":"Front. Neurosci"},{"key":"B71","doi-asserted-by":"publisher","first-page":"455","DOI":"10.3390\/e24040455","article-title":"Robust spike-based continual meta-learning improved by restricted minimum error entropy criterion","volume":"24","author":"Yang","year":"","journal-title":"Entropy"},{"key":"B72","doi-asserted-by":"publisher","first-page":"e1006227","DOI":"10.1371\/journal.pcbi.1006227","article-title":"General differential Hebbian learning: capturing temporal relations between events in neural networks and the brain","volume":"14","author":"Zappacosta","year":"2018","journal-title":"PLoS Comput. Biol"},{"key":"B73","doi-asserted-by":"publisher","first-page":"660","DOI":"10.1126\/science.1254126","article-title":"Long-range and local circuits for top-down modulation of visual cortex processing","volume":"345","author":"Zhang","year":"2014","journal-title":"Science"}],"container-title":["Frontiers in Computational Neuroscience"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fncom.2023.1136010\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,5,24]],"date-time":"2023-05-24T06:09:29Z","timestamp":1684908569000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fncom.2023.1136010\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,5,24]]},"references-count":73,"alternative-id":["10.3389\/fncom.2023.1136010"],"URL":"https:\/\/doi.org\/10.3389\/fncom.2023.1136010","relation":{},"ISSN":["1662-5188"],"issn-type":[{"value":"1662-5188","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,5,24]]},"article-number":"1136010"}}