{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,18]],"date-time":"2026-09-18T21:03:30Z","timestamp":1789765410201,"version":"4.0.1"},"reference-count":71,"publisher":"MIT Press","issue":"10","content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2026,9,18]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Working memory requires the brain to maintain information from the recent past to guide ongoing behavior. Neurons can contribute to this capacity by slowly integrating their inputs over time, creating persistent activity that outlasts the original stimulus. However, when these slowly integrating neurons are organized hierarchically, they introduce cumulative delays that create a fundamental challenge for learning: teaching signals that indicate whether behavior was correct or incorrect arrive out-of-sync with the neural activity they are meant to instruct. Here, we demonstrate that neurons enhanced with an adaptive current can compensate for these delays by responding to external stimuli prospectively, effectively predicting future inputs to synchronize with them. First, we show that such prospective neurons enable teaching signal synchronization across a range of learning algorithms that propagate error signals through hierarchical networks. Second, we demonstrate that this successfully guides learning in slowly integrating neurons, enabling the formation and retrieval of memories over extended timescales. We support our findings with a mathematical analysis of the prospective coding mechanism and learning experiments on motor control tasks. Together, our results reveal how neural adaptation could solve a critical timing problem and enable efficient learning in dynamic environments.<\/jats:p>","DOI":"10.1162\/neco.a.1571","type":"journal-article","created":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T23:20:26Z","timestamp":1786749626000},"page":"1785-1828","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":0,"title":["Teaching Signal Synchronization in Deep Neural Networks With Prospective Neurons"],"prefix":"10.1162","volume":"38","author":[{"given":"Nicolas","family":"Zucchet","sequence":"first","affiliation":[{"name":"Stanford University, Stanford, CA 94305, U.S.A. nzucchet@stanford.edu"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qianqian","family":"Feng","sequence":"additional","affiliation":[{"name":"Gatsby Computational Neuroscience Unit, University College London, London W1T 4JG, U.K. qianqian.feng@ucl.ac.uk"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Axel","family":"Laborieux","sequence":"additional","affiliation":[{"name":"Friedrich Miescher Institute for Biomedical Research, 4056 Basel, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Friedemann","family":"Zenke","sequence":"additional","affiliation":[{"name":"Friedrich Miescher Institute for Biomedical Research, 4056 Basel, Switzerland; and Faculty of Science, University of Basel, 4001 Basel, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Walter","family":"Senn","sequence":"additional","affiliation":[{"name":"Department of Physiology, University of Bern, 3012 Bern, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jo\u00e3o","family":"Sacramento","sequence":"additional","affiliation":[{"name":"Google, Paradigms of Intelligence Team"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"281","published-online":{"date-parts":[[2026,9,18]]},"reference":[{"issue":"5\u20136","key":"2026091816155631000_bib1","doi-asserted-by":"publisher","DOI":"10.1016\/S0361-9230(99)00161-6","article-title":"Lapicque\u2019s introduction of the integrate-and-fire model neuron (1907)","volume":"50","author":"Abbott","year":"1999","journal-title":"Brain Research Bulletin"},{"issue":"1","key":"2026091816155631000_bib2","article-title":"A learning algorithm for Boltzmann machines","volume":"9","author":"Ackley","year":"1985","journal-title":"Cognitive Science"},{"key":"2026091816155631000_bib3","volume-title":"Foundations of analog and digital electronic circuits","author":"Agarwal","year":"2005"},{"key":"2026091816155631000_bib4","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-642-83740-1_22","article-title":"Backpropagation in perceptrons with feedback","volume-title":"Neural computers","author":"Almeida","year":"1989"},{"issue":"4","key":"2026091816155631000_bib5","doi-asserted-by":"publisher","DOI":"10.1162\/neco_a_01648","article-title":"Frequency propagation: Multi-mechanism learning in nonlinear physical networks","volume":"36","author":"Anisetti","year":"2024","journal-title":"Neural Computation"},{"key":"2026091816155631000_bib6","article-title":"Deep equilibrium models","volume-title":"Advances in neural information processing systems","author":"Bai","year":"2019"},{"issue":"4","key":"2026091816155631000_bib7","doi-asserted-by":"crossref","DOI":"10.1162\/neco.1991.3.4.526","article-title":"Contrastive learning and neural oscillations","volume":"3","author":"Baldi","year":"1991","journal-title":"Neural Computation"},{"issue":"1","key":"2026091816155631000_bib8","doi-asserted-by":"publisher","first-page":"3625","DOI":"10.1038\/s41467-020-17236-y","article-title":"A solution to the learning dilemma for recurrent networks of spiking neurons","volume":"11","author":"Bellec","year":"2020","journal-title":"Nature Communications"},{"issue":"11","key":"2026091816155631000_bib9","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2022.3153985","article-title":"Online spatio-temporal learning in deep neural networks","volume":"34","author":"Bohnstingl","year":"2023","journal-title":"IEEE Transactions on Neural Networks and Learning Systems"},{"key":"2026091816155631000_bib10","author":"Brandt","year":"2024","journal-title":"Prospective and retrospective coding in cortical neurons."},{"issue":"5","key":"2026091816155631000_bib11","doi-asserted-by":"publisher","DOI":"10.1152\/jn.00686.2005","article-title":"Adaptive exponential integrate-and-fire model as an effective description of neuronal activity","volume":"94","author":"Brette","year":"2005","journal-title":"Journal of Neurophysiology"},{"key":"2026091816155631000_bib12","doi-asserted-by":"crossref","DOI":"10.1007\/978-0-387-87821-8","volume-title":"Implicit functions and solution mappings","author":"Dontchev","year":"2009"},{"issue":"1","key":"2026091816155631000_bib13","doi-asserted-by":"publisher","DOI":"10.1162\/089976600300015961","article-title":"Reinforcement learning in continuous time and space","volume":"12","author":"Doya","year":"2000","journal-title":"Neural Computation"},{"issue":"66","key":"2026091816155631000_bib14","article-title":"Backpropagation through space, time, and the brain","volume":"17","author":"Ellenberger","year":"2026","journal-title":"Nature Communications"},{"key":"2026091816155631000_bib15","doi-asserted-by":"crossref","DOI":"10.1017\/CBO9781107447615","volume-title":"Neuronal dynamics: From single neurons to networks and models of cognition","author":"Gerstner","year":"2014"},{"key":"2026091816155631000_bib16","doi-asserted-by":"publisher","DOI":"10.3389\/fncir.2018.00053","article-title":"Eligibility traces and plasticity on behavioral time scales: Experimental support of neoHebbian three-factor learning rules","volume":"12","author":"Gerstner","year":"2018","journal-title":"Frontiers in Neural Circuits"},{"key":"2026091816155631000_bib17","doi-asserted-by":"publisher","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"},{"issue":"1","key":"2026091816155631000_bib18","doi-asserted-by":"publisher","DOI":"10.1111\/j.1551-6708.1987.tb00862.x","article-title":"Competitive learning: From interactive activation to adaptive resonance","volume":"11","author":"Grossberg","year":"1987","journal-title":"Cognitive Science"},{"key":"2026091816155631000_bib19","article-title":"Efficiently modeling long sequences with structured state spaces","volume-title":"Proceedings of the International Conference on Learning Representations","author":"Gu","year":"2022"},{"key":"2026091816155631000_bib20","article-title":"Latent equilibrium: A unified learning theory for arbitrarily fast computation with arbitrarily slow neurons","volume-title":"Advances in neural information processing systems, 34","author":"Haider","year":"2021"},{"issue":"4","key":"2026091816155631000_bib21","doi-asserted-by":"publisher","DOI":"10.1113\/jphysiol.1952.sp004764","article-title":"A quantitative description of membrane current and its application to conduction and excitation in nerve","volume":"117","author":"Hodgkin","year":"1952","journal-title":"Journal of Physiology"},{"issue":"6\u20137","key":"2026091816155631000_bib22","doi-asserted-by":"publisher","DOI":"10.1016\/S0893-6080(01)00078-8","article-title":"Resonate-and-fire neurons","volume":"14","author":"Izhikevich","year":"2001","journal-title":"Neural Networks"},{"issue":"10","key":"2026091816155631000_bib23","doi-asserted-by":"publisher","DOI":"10.1093\/cercor\/bhl152","article-title":"Solving the distal reward problem through linkage of STDP and dopamine signaling","volume":"17","author":"Izhikevich","year":"2007","journal-title":"Cerebral Cortex"},{"issue":"Series D","key":"2026091816155631000_bib24","article-title":"A new approach to linear filtering and prediction problems","volume":"82","author":"Kalman","year":"1960","journal-title":"Transactions of the ASME-Journal of Basic Engineering"},{"key":"2026091816155631000_bib25","volume-title":"Principles of neural science","author":"Kandel","year":"2000"},{"issue":"6","key":"2026091816155631000_bib26","doi-asserted-by":"crossref","DOI":"10.1085\/jgp.59.6.734","article-title":"Dynamics of encoding in a population of neurons","volume":"59","author":"Knight","year":"1972","journal-title":"Journal of General Physiology"},{"key":"2026091816155631000_bib27","doi-asserted-by":"crossref","DOI":"10.52202\/068431-0941","article-title":"Holomorphic equilibrium propagation computes exact gradients through finite size oscillations","volume-title":"Advances in neural information processing systems","author":"Laborieux","year":"2022"},{"key":"2026091816155631000_bib28","article-title":"Improving equilibrium propagation without weight symmetry through Jacobian homeostasis","volume-title":"Proceedings of the International Conference on Learning Representations","author":"Laborieux","year":"2024"},{"key":"2026091816155631000_bib29","doi-asserted-by":"crossref","DOI":"10.1007\/978-0-85729-664-1","volume-title":"Adaptive control: Algorithms, analysis and applications","author":"Landau","year":"2011"},{"key":"2026091816155631000_bib30","article-title":"gymnax: A JAX-based reinforcement learning environment library","author":"Lange","year":"2022"},{"key":"2026091816155631000_bib31","article-title":"A theoretical framework for back-propagation","author":"LeCun","year":"1988","journal-title":"Proceedings of the 1998 Connectionist Models Summer School"},{"key":"2026091816155631000_bib32","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-319-23528-8_31","volume-title":"Difference target propagation.","author":"Lee","year":"2015"},{"issue":"1","key":"2026091816155631000_bib33","doi-asserted-by":"publisher","DOI":"10.1038\/ncomms13276","article-title":"Random synaptic feedback weights support error backpropagation for deep learning","volume":"7","author":"Lillicrap","year":"2016","journal-title":"Nature Communications"},{"issue":"3","key":"2026091816155631000_bib34","doi-asserted-by":"crossref","DOI":"10.1016\/j.cosrev.2009.03.005","article-title":"Reservoir computing approaches to recurrent neural network training","volume":"3","author":"Luko\u0161evi\u010dius","year":"2009","journal-title":"Computer Science Review"},{"key":"2026091816155631000_bib44","doi-asserted-by":"crossref","DOI":"10.1017\/S0962492900002269","article-title":"Numerical methods for differential algebraic equations","volume":"1","author":"M\u00e4rz","year":"1992","journal-title":"Acta Numerica"},{"key":"2026091816155631000_bib35","doi-asserted-by":"crossref","DOI":"10.52202\/068431-2435","article-title":"The least-control principle for local learning at equilibrium","author":"Meulemans","year":"2022","journal-title":"NeurIPS 2022"},{"key":"2026091816155631000_bib36","article-title":"Spike frequency adaptation implements anticipative tracking in continuous attractor neural networks","volume-title":"Advances in neural information processing systems","author":"Mi","year":"2014"},{"issue":"8","key":"2026091816155631000_bib37","doi-asserted-by":"publisher","DOI":"10.1016\/S0893-6080(96)00035-4","article-title":"Forward models for physiological motor control","volume":"9","author":"Miall","year":"1996","journal-title":"Neural Networks"},{"issue":"1","key":"2026091816155631000_bib38","doi-asserted-by":"publisher","DOI":"10.1016\/j.tins.2022.09.007","article-title":"Where is the error? Hierarchical predictive coding through dendritic error computation","volume":"46","author":"Mikulasch","year":"2023","journal-title":"Trends in Neurosciences"},{"issue":"2","key":"2026091816155631000_bib39","doi-asserted-by":"publisher","DOI":"10.1111\/j.1559-3584.1922.tb04958.x","article-title":"Directional stability of automatically steered bodies","volume":"34","author":"Minorsky","year":"1922","journal-title":"Journal of the American Society for Naval Engineers"},{"key":"2026091816155631000_bib40","article-title":"Asynchronous methods for deep reinforcement learning","author":"Mnih","year":"2016","journal-title":"Proceedings of the International Conference on Machine Learning"},{"key":"2026091816155631000_bib41","doi-asserted-by":"crossref","DOI":"10.1016\/B978-1-4832-1448-1.50007-X","article-title":"Contrastive Hebbian learning in the continuous Hopfield model","author":"Movellan","year":"1991","journal-title":"Connectionist Models"},{"key":"2026091816155631000_bib42","article-title":"A focused backpropagation algorithm for temporal pattern recognition","volume":"3","author":"Mozer","year":"1989","journal-title":"Complex Systems"},{"key":"2026091816155631000_bib43","doi-asserted-by":"crossref","DOI":"10.7554\/eLife.43299","article-title":"Local online learning in recurrent networks with random feedback","volume":"8","author":"Murray","year":"2019","journal-title":"eLife"},{"key":"2026091816155631000_bib45","article-title":"Direct feedback alignment provides learning in deep neural networks","volume-title":"Advances in neural information processing systems","author":"N\u00f8kland","year":"2016"},{"key":"2026091816155631000_bib46","article-title":"Resurrecting recurrent neural networks for long sequences","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Orvieto","year":"2023"},{"issue":"2","key":"2026091816155631000_bib47","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1989.1.2.161","article-title":"Recurrent backpropagation and the dynamical approach to adaptive neural computation","volume":"1","author":"Pineda","year":"1989","journal-title":"Neural Computation"},{"key":"2026091816155631000_bib48","article-title":"Introduction to optimization","author":"Polyak","year":"1987"},{"key":"2026091816155631000_bib49","doi-asserted-by":"crossref","DOI":"10.1007\/s10513-005-0132-z","article-title":"Gradient methods for nonstationary unconstrained optimization problems","volume":"66","author":"Popkov","year":"2005","journal-title":"Automation and Remote Control"},{"key":"2026091816155631000_bib50","doi-asserted-by":"publisher","DOI":"10.1016\/j.conb.2018.08.003","article-title":"Dendritic solutions to the credit assignment problem","volume":"54","author":"Richards","year":"2019","journal-title":"Current Opinion in Neurobiology"},{"issue":"11","key":"2026091816155631000_bib51","doi-asserted-by":"publisher","DOI":"10.1038\/s41593-019-0520-2","article-title":"A deep learning framework for neuroscience","volume":"22","author":"Richards","year":"2019","journal-title":"Nature Neuroscience"},{"issue":"6088","key":"2026091816155631000_bib52","doi-asserted-by":"publisher","DOI":"10.1038\/323533a0","article-title":"Learning representations by back-propagating errors","volume":"323","author":"Rumelhart","year":"1986","journal-title":"Nature"},{"key":"2026091816155631000_bib53","article-title":"Dendritic cortical microcircuits approximate the backpropagation algorithm","volume-title":"Advances in neural information processing systems","author":"Sacramento","year":"2018"},{"key":"2026091816155631000_bib54","doi-asserted-by":"publisher","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":"Frontiers in Computational Neuroscience"},{"key":"2026091816155631000_bib55","article-title":"A complex-valued firing-rate model that approximates the dynamics of spiking networks","volume":"10","author":"Schaffer","year":"2013","journal-title":"PLOS Computational Biology"},{"key":"2026091816155631000_bib56","doi-asserted-by":"crossref","DOI":"10.7554\/eLife.89674.3.sa3","article-title":"A neuronal least-action principle for real-time learning in cortical circuits","author":"Senn","year":"2024","journal-title":"eLife"},{"issue":"20","key":"2026091816155631000_bib57","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2017.2728498","article-title":"Prediction-correction algorithms for time-varying constrained optimization","volume":"65","author":"Simonetto","year":"2017","journal-title":"IEEE Transactions on Signal Processing"},{"issue":"11","key":"2026091816155631000_bib58","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2020.3003156","article-title":"Time-varying convex optimization: Time-structured algorithms and applications","volume":"108","author":"Simonetto","year":"2020","journal-title":"Proceedings of the IEEE"},{"key":"2026091816155631000_bib59","article-title":"Making deep Q-learning methods robust to time discretization","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Tallec","year":"2019"},{"issue":"6","key":"2026091816155631000_bib60","doi-asserted-by":"publisher","DOI":"10.1152\/jn.2002.87.6.2753","article-title":"Dendritic resonance in rat neocortical pyramidal cells","volume":"87","author":"Ulrich","year":"2002","journal-title":"Journal of Neurophysiology"},{"issue":"5293","key":"2026091816155631000_bib61","doi-asserted-by":"publisher","DOI":"10.1126\/science.274.5293.1724","article-title":"Chaos in neuronal networks with balanced excitatory and inhibitory activity","volume":"274","author":"van Vreeswijk","year":"1996","journal-title":"Science"},{"issue":"6","key":"2026091816155631000_bib62","article-title":"Chaotic balanced state in a model of cortical circuits","volume":"10","author":"van Vreeswijk","year":"1998","journal-title":"Neural Computation"},{"key":"2026091816155631000_bib63","author":"Werbos","year":"1974","journal-title":"Beyond regression: New tools for prediction and analysis in the behavioral sciences"},{"issue":"5","key":"2026091816155631000_bib64","doi-asserted-by":"publisher","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 Computation"},{"issue":"2","key":"2026091816155631000_bib65","doi-asserted-by":"publisher","first-page":"270","DOI":"10.1162\/neco.1989.1.2.270","article-title":"A learning algorithm for continually running fully recurrent neural networks","volume":"1","author":"Williams","year":"1989","journal-title":"Neural Computation"},{"issue":"9","key":"2026091816155631000_bib66","doi-asserted-by":"publisher","DOI":"10.1016\/S1364-6613(98)01221-2","article-title":"Internal models in the cerebellum","volume":"2","author":"Wolpert","year":"1998","journal-title":"Trends in Cognitive Sciences"},{"issue":"6","key":"2026091816155631000_bib67","doi-asserted-by":"publisher","DOI":"10.1162\/neco_a_01086","article-title":"Superspike: Supervised learning in multilayer spiking neural networks","volume":"30","author":"Zenke","year":"2018","journal-title":"Neural Computation"},{"key":"2026091816155631000_bib68","doi-asserted-by":"crossref","DOI":"10.1109\/CCECE.1998.685646","article-title":"A novel technique for tracking time-varying minimum and its applications","volume":"2","author":"Zhao","year":"1998","journal-title":"Proceedings of the IEEE Canadian Conference on Electrical and Computer Engineering"},{"key":"2026091816155631000_bib69","article-title":"Online convex programming and generalized infinitesimal gradient ascent","volume-title":"Proceedings of the International Conference on Machine Learning","author":"Zinkevich","year":"2003"},{"key":"2026091816155631000_bib70","doi-asserted-by":"crossref","DOI":"10.52202\/075280-0460","article-title":"Online learning of long-range dependencies","volume-title":"Advances in neural information processing systems","author":"Zucchet","year":"2023"},{"issue":"12","key":"2026091816155631000_bib71","doi-asserted-by":"publisher","DOI":"10.1162\/neco_a_01547","article-title":"Beyond backpropagation: Implicit gradients for bilevel optimization","volume":"34","author":"Zucchet","year":"2022","journal-title":"Neural Computation"}],"container-title":["Neural Computation"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/38\/10\/1785\/2620904\/neco.a.1571.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/38\/10\/1785\/2620904\/neco.a.1571.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,18]],"date-time":"2026-09-18T20:16:25Z","timestamp":1789762585000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/neco\/article\/38\/10\/1785\/138264\/Teaching-Signal-Synchronization-in-Deep-Neural"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,18]]},"references-count":71,"journal-issue":{"issue":"10","published-online":{"date-parts":[[2026,9,18]]},"published-print":{"date-parts":[[2026,9,18]]}},"URL":"https:\/\/doi.org\/10.1162\/neco.a.1571","relation":{},"ISSN":["0899-7667","1530-888X"],"issn-type":[{"value":"0899-7667","type":"print"},{"value":"1530-888X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,10]]},"published":{"date-parts":[[2026,9,18]]}}}