{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T23:12:47Z","timestamp":1763507567061,"version":"3.45.0"},"reference-count":61,"publisher":"MIT Press","issue":"12","content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2025,11,18]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Neural manifolds are an attractive theoretical framework for characterizing the complex behaviors of neural populations. However, many of the tools for identifying these low-dimensional subspaces are correlational and provide limited insight into the underlying dynamics. The ability to precisely control the latent activity of a circuit would allow researchers to investigate the structure and function of neural manifolds. We simulate controlling the latent dynamics of a neural population using closed-loop, dynamically generated sensory inputs. Using a spiking neural network (SNN) as a model of a neural circuit, we find low-dimensional representations of both the network activity (the neural manifold) and a set of salient visual stimuli. The fields of classical and optimal control offer a range of methods to choose from for controlling dynamics on the neural manifold, which differ in performance, computational cost, and ease of implementation. Here, we focus on two commonly used control methods: proportional-integral-derivative (PID) control and model predictive control (MPC). PID is a computationally lightweight controller that is simple to implement. In contrast, MPC is a model-based, anticipatory controller with a much higher computational cost and engineering overhead. We evaluate both methods on trajectory-following tasks in latent space, under partial observability and in the presence of unknown noise. While both controllers in some cases were able to successfully control the latent dynamics on the neural manifold, MPC consistently produced more accurate control and required less hyperparameter tuning. These results demonstrate how MPC can be applied on the neural manifold using data-driven dynamics models and provide a framework to experimentally test for causal relationships between manifold dynamics and external stimuli.<\/jats:p>","DOI":"10.1162\/neco.a.37","type":"journal-article","created":{"date-parts":[[2025,10,29]],"date-time":"2025-10-29T18:46:07Z","timestamp":1761763567000},"page":"2125-2157","update-policy":"https:\/\/linproxy.fan.workers.dev:443\/https\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":0,"title":["Model Predictive Control on the Neural Manifold"],"prefix":"10.1162","volume":"37","author":[{"given":"Christof","family":"Fehrman","sequence":"first","affiliation":[{"name":"Department of Mechanical Engineering and Materials Science, Duke University, Durham NC 27708, USA christof.fehrman@duke.edu"}]},{"given":"C. Daniel","family":"Meliza","sequence":"additional","affiliation":[{"name":"Department of Psychology and Program in Fundamental Neuroscience, University of Virginia, Charlottesville VA 22904, USA cdm8j@virginia.edu"}]}],"member":"281","published-online":{"date-parts":[[2025,11,18]]},"reference":[{"issue":"1","key":"2025111818105543700_bib1","doi-asserted-by":"publisher","first-page":"85","DOI":"10.1038\/s41551-023-01106-1","article-title":"Dynamical flexible inference of nonlinear latent factors and structures in neural population activity","volume":"8","author":"Abbaspourazad","year":"2024","journal-title":"Nature Biomedical Engineering"},{"issue":"10","key":"2025111818105543700_bib2","doi-asserted-by":"publisher","first-page":"1356","DOI":"10.1038\/s41593-021-00902-9","article-title":"Probing neural codes with two-photon holographic optogenetics","volume":"24","author":"Adesnik","year":"2021","journal-title":"Nature Neuroscience"},{"issue":"1","key":"2025111818105543700_bib3","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s12532-018-0139-4","article-title":"CasADi: A software framework for nonlinear optimization and optimal control","volume":"11","author":"Andersson","year":"2019","journal-title":"Mathematical Programming Computation"},{"journal-title":"Deep model predictive control with online learning for complex physical systems","year":"2019","author":"Bieker","key":"2025111818105543700_bib4"},{"issue":"3","key":"2025111818105543700_bib5","doi-asserted-by":"publisher","first-page":"036006","DOI":"10.1088\/1741-2552\/abb89c","article-title":"State-space optimal feedback control of optogenetically driven neural activity","volume":"18","author":"Bolus","year":"2021","journal-title":"Journal of Neural Engineering"},{"key":"2025111818105543700_bib6","doi-asserted-by":"crossref","DOI":"10.1017\/9781108380690","author":"Brunton","year":"2019","journal-title":"Data-driven science and engineering: Machine learning, dynamical systems, and control"},{"issue":"9","key":"2025111818105543700_bib7","doi-asserted-by":"publisher","first-page":"1512","DOI":"10.1038\/s41593-019-0460-x","article-title":"The intrinsic attractor manifold and population dynamics of a canonical cognitive circuit across waking and sleep","volume":"22","author":"Chaudhuri","year":"2019","journal-title":"Nature Neuroscience"},{"issue":"7","key":"2025111818105543700_bib8","doi-asserted-by":"publisher","first-page":"433","DOI":"10.1038\/s43588-022-00281-6","article-title":"Automated discovery of fundamental variables hidden in experimental data","volume":"2","author":"Chen","year":"2022","journal-title":"Nature Computational Science"},{"key":"2025111818105543700_bib9","doi-asserted-by":"publisher","first-page":"137","DOI":"10.1016\/j.conb.2021.10.010","article-title":"Neural population geometry: An approach for understanding biological and artificial neural networks","volume":"70","author":"Chung","year":"2021","journal-title":"Current Opinion in Neurobiology"},{"issue":"7","key":"2025111818105543700_bib10","doi-asserted-by":"publisher","first-page":"1545","DOI":"10.1162\/neco_a_01515","article-title":"Reduced-dimension, biophysical neuron models constructed from observed data","volume":"34","author":"Clark","year":"2022","journal-title":"Neural Computation"},{"issue":"11","key":"2025111818105543700_bib11","doi-asserted-by":"publisher","first-page":"1500","DOI":"10.1038\/nn.3776","article-title":"Dimensionality reduction for large-scale neural recordings","volume":"17","author":"Cunningham","year":"2014","journal-title":"Nature Neuroscience"},{"issue":"4","key":"2025111818105543700_bib12","doi-asserted-by":"publisher","first-page":"337","DOI":"10.1038\/s41551-022-00962-7","article-title":"Aligning latent representations of neural activity","volume":"7","author":"Dabagia","year":"2023","journal-title":"Nature Biomedical Engineering"},{"issue":"49","key":"2025111818105543700_bib13","doi-asserted-by":"publisher","first-page":"16601","DOI":"10.1523\/JNEUROSCI.2770-10.2010","article-title":"Expectation and surprise determine neural population responses in the ventral visual stream","volume":"30","author":"Egner","year":"2010","journal-title":"Journal of Neuroscience"},{"issue":"9","key":"2025111818105543700_bib14","doi-asserted-by":"publisher","first-page":"1016","DOI":"10.1109\/JPROC.2023.3308088","article-title":"Training spiking neural networks using lessons from deep learning","volume":"111","author":"Eshraghian","year":"2023","journal-title":"Proceedings of the IEEE"},{"issue":"5","key":"2025111818105543700_bib15","doi-asserted-by":"publisher","DOI":"10.1088\/1741-2552\/ad731f","article-title":"Nonlinear model predictive control of a conductance-based neuron model via data-driven forecasting","volume":"21","author":"Fehrman","year":"2024","journal-title":"Journal of Neural Engineering"},{"key":"2025111818105543700_bib16","doi-asserted-by":"publisher","DOI":"10.1016\/j.conengprac.2023.105676","article-title":"do-mpc: Towards FAIR nonlinear and robust model predictive control","volume":"140","author":"Fiedler","year":"2023","journal-title":"Control Engineering Practice"},{"key":"2025111818105543700_bib17","doi-asserted-by":"publisher","DOI":"10.5772\/14183","article-title":"Hidden Markov models in the neurosciences","author":"Florian","year":"2011","journal-title":"Hidden Markov models, theory and applications"},{"key":"2025111818105543700_bib18","doi-asserted-by":"publisher","DOI":"10.1101\/2023.07.18.549575","author":"Fortunato","year":"2023","journal-title":"Nonlinear manifolds underlie neural population activity during behaviour"},{"issue":"9","key":"2025111818105543700_bib19","doi-asserted-by":"publisher","first-page":"1195","DOI":"10.1016\/j.conengprac.2004.10.008","article-title":"Feedback control of Hodgkin\u2013Huxley nerve cell dynamics","volume":"13","author":"Fr\u00f6hlich","year":"2005","journal-title":"Control Engineering Practice"},{"issue":"5","key":"2025111818105543700_bib20","doi-asserted-by":"publisher","first-page":"978","DOI":"10.1016\/j.neuron.2017.05.025","article-title":"Neural manifolds for the control of movement","volume":"94","author":"Gallego","year":"2017","journal-title":"Neuron"},{"issue":"1","key":"2025111818105543700_bib21","doi-asserted-by":"publisher","first-page":"2450006","DOI":"10.1142\/S0129065724500060","article-title":"Unsupervised neural manifold alignment for stable decoding of movement from cortical signals","volume":"34","author":"Ganjali","year":"2024","journal-title":"International Journal of Neural Systems"},{"key":"2025111818105543700_bib22","doi-asserted-by":"publisher","first-page":"e67855","DOI":"10.7554\/eLife.67855","article-title":"Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires","volume":"10","author":"Goffinet","year":"2021","journal-title":"eLife"},{"issue":"1","key":"2025111818105543700_bib23","doi-asserted-by":"publisher","first-page":"645","DOI":"10.1038\/s42003-022-03579-3","article-title":"Variational autoencoders learn transferrable representations of metabolomics data","volume":"5","author":"Gomari","year":"2022","journal-title":"Communications Biology"},{"key":"2025111818105543700_bib24","first-page":"1","article-title":"MARBLE: Interpretable representations of neural population dynamics using geometric deep learning","author":"Gosztolai","year":"2025","journal-title":"Nature Methods"},{"journal-title":"The organization of behavior","year":"1949","author":"Hebb","key":"2025111818105543700_bib25"},{"issue":"1","key":"2025111818105543700_bib26","doi-asserted-by":"publisher","first-page":"269","DOI":"10.1146\/annurev-control-090419-075625","article-title":"Learning- based model predictive control: Toward safe learning in control","volume":"3","author":"Hewing","year":"2020","journal-title":"Annual Review of Control, Robotics, and Autonomous Systems"},{"key":"2025111818105543700_bib27","volume":"1","author":"James","year":"1890","journal-title":"The principles of psychology"},{"key":"2025111818105543700_bib28","volume":"2","author":"Johnson","year":"2016","journal-title":"Structured VAEs: Composing probabilistic graphical models and variational autoencoders."},{"issue":"2219","key":"2025111818105543700_bib29","doi-asserted-by":"publisher","first-page":"20180335","DOI":"10.1098\/rspa.2018.0335","article-title":"Sparse identification of nonlinear dynamics for model predictive control in the low-data limit","volume":"474","author":"Kaiser","year":"2018","journal-title":"Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences"},{"issue":"3","key":"2025111818105543700_bib30","doi-asserted-by":"publisher","DOI":"10.1088\/2632-2153\/abf0f5","article-title":"Data-driven discovery of Koopman eigenfunctions for control","volume":"2","author":"Kaiser","year":"2021","journal-title":"Machine Learning: Science and Technology"},{"issue":"12","key":"2025111818105543700_bib31","doi-asserted-by":"publisher","first-page":"1572","DOI":"10.1038\/s41592-022-01675-0","article-title":"A large-scale neural network training framework for generalized estimation of single-trial population dynamics","volume":"19","author":"Keshtkaran","year":"2022","journal-title":"Nature Methods"},{"issue":"12","key":"2025111818105543700_bib32","doi-asserted-by":"publisher","first-page":"744","DOI":"10.1038\/s41583-022-00642-0","article-title":"Attractor and integrator networks in the brain","volume":"23","author":"Khona","year":"2022","journal-title":"Nature Reviews Neuroscience"},{"issue":"6340","key":"2025111818105543700_bib33","doi-asserted-by":"publisher","first-page":"849","DOI":"10.1126\/science.aal4835","article-title":"Ring attractor dynamics in the drosophila central brain","volume":"356","author":"Kim","year":"2017","journal-title":"Science"},{"journal-title":"Auto-encoding variational Bayes","year":"2013","author":"Kingma","key":"2025111818105543700_bib34"},{"issue":"6","key":"2025111818105543700_bib35","doi-asserted-by":"publisher","first-page":"363","DOI":"10.1038\/s41583-023-00693-x","article-title":"A unifying perspective on neural manifolds and circuits for cognition","volume":"24","author":"Langdon","year":"2023","journal-title":"Nature Reviews Neuroscience"},{"issue":"11","key":"2025111818105543700_bib36","doi-asserted-by":"publisher","first-page":"2278","DOI":"10.1109\/5.726791","article-title":"Gradient-based learning applied to document recognition","volume":"86","author":"LeCun","year":"1998","journal-title":"Proceedings of the IEEE"},{"key":"2025111818105543700_bib37","article-title":"Maximum likelihood estimation of intrinsic dimension","volume":"17","author":"Levina","year":"2004","journal-title":"Advances in neural information processing systems"},{"key":"2025111818105543700_bib38","first-page":"914","article-title":"Bayesian learning and inference in recurrent switching linear dynamical systems","volume-title":"Proceedings of the 20th International Conference on Artificial Intelligence and Statistics","author":"Linderman","year":"2017"},{"issue":"1","key":"2025111818105543700_bib39","doi-asserted-by":"publisher","first-page":"4950","DOI":"10.1038\/s41467-018-07210-0","article-title":"Deep learning for universal linear embeddings of nonlinear dynamics","volume":"9","author":"Lusch","year":"2018","journal-title":"Nature Communications"},{"key":"2025111818105543700_bib40","doi-asserted-by":"crossref","first-page":"275","DOI":"10.1142\/9781848162778_0008","article-title":"Liquid state machines: Motivation theory, and applications","author":"Maass","year":"2011","journal-title":"Computability in context: Computation and logic in the real world"},{"issue":"7474","key":"2025111818105543700_bib41","doi-asserted-by":"publisher","first-page":"78","DOI":"10.1038\/nature12742","article-title":"Context-dependent computation by recurrent dynamics in prefrontal cortex","volume":"503","author":"Mante","year":"2013","journal-title":"Nature"},{"issue":"6","key":"2025111818105543700_bib42","doi-asserted-by":"publisher","first-page":"51","DOI":"10.1109\/MSP.2019.2931595","article-title":"Surrogate gradient learning in spiking neural networks: Bringing the power of gradient-based optimization to spiking neural networks","volume":"36","author":"Neftci","year":"2019","journal-title":"IEEE Signal Processing Magazine"},{"issue":"10","key":"2025111818105543700_bib43","doi-asserted-by":"publisher","first-page":"805","DOI":"10.1038\/s41592-018-0109-9","article-title":"Inferring single-trial neural population dynamics using sequential auto-encoders","volume":"15","author":"Pandarinath","year":"2018","journal-title":"Nature Methods"},{"issue":"14","key":"2025111818105543700_bib44","doi-asserted-by":"publisher","first-page":"R656","DOI":"10.1016\/j.cub.2016.05.029","article-title":"Dimensionality reduction in neuroscience","volume":"26","author":"Pang","year":"2016","journal-title":"Current Biology"},{"key":"2025111818105543700_bib45","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-319-77489-3","author":"Rakovi\u00e9","year":"2019","journal-title":"Handbook of model predictive control"},{"issue":"4","key":"2025111818105543700_bib46","doi-asserted-by":"publisher","first-page":"2397","DOI":"10.1109\/LRA.2023.3246839","article-title":"Real-time neural MPC: Deep learning model predictive control for quadrotors and agile robotic platforms","volume":"8","author":"Salzmann","year":"2023","journal-title":"IEEE Robotics and Automation Letters"},{"issue":"1","key":"2025111818105543700_bib47","doi-asserted-by":"publisher","first-page":"140","DOI":"10.1038\/s41593-020-00733-0","article-title":"Modeling behaviorally relevant neural dynamics enabled by preferential subspace identification","volume":"24","author":"Sani","year":"2021","journal-title":"Nature Neuroscience"},{"issue":"7960","key":"2025111818105543700_bib48","doi-asserted-by":"publisher","first-page":"360","DOI":"10.1038\/s41586-023-06031-6","article-title":"Learnable latent embeddings for joint behavioural and neural analysis","volume":"617","author":"Schneider","year":"2023","journal-title":"Nature"},{"issue":"5\u20136","key":"2025111818105543700_bib49","doi-asserted-by":"publisher","first-page":"1327","DOI":"10.1007\/s00170-021-07682-3","article-title":"Review on model predictive control: An engineering perspective","volume":"117","author":"Schwenzer","year":"2021","journal-title":"International Journal of Advanced Manufacturing Technology"},{"issue":"4","key":"2025111818105543700_bib50","doi-asserted-by":"publisher","first-page":"373","DOI":"10.1515\/bmt-2021-0302","article-title":"Active fault tolerant deep brain stimulator for epilepsy using deep neural network","volume":"68","author":"Senthilvelmurugan","year":"2023"},{"issue":"6","key":"2025111818105543700_bib51","doi-asserted-by":"publisher","first-page":"066019","DOI":"10.1088\/1741-2552\/ac9b94","article-title":"Modeling and inference methods for switching regime-dependent dynamical systems with multiscale neural observations","volume":"19","author":"Song","year":"2022","journal-title":"Journal of Neural Engineering"},{"issue":"6539","key":"2025111818105543700_bib52","doi-asserted-by":"publisher","first-page":"eabf4588","DOI":"10.1126\/science.abf4588","article-title":"Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings","volume":"372","author":"Steinmetz","year":"2021","journal-title":"Science"},{"issue":"6268","key":"2025111818105543700_bib53","doi-asserted-by":"publisher","first-page":"734","DOI":"10.1038\/344734a0","article-title":"Nonlinear forecasting as a way of distinguishing chaos from measurement error in time series","volume":"344","author":"Sugihara","year":"1990","journal-title":"Nature"},{"issue":"3","key":"2025111818105543700_bib54","doi-asserted-by":"publisher","first-page":"626","DOI":"10.1162\/NECO_a_00409","article-title":"Opening the black box: Low-dimensional dynamics in high-dimensional recurrent neural networks","volume":"25","author":"Sussillo","year":"2013","journal-title":"Neural Computation"},{"issue":"7","key":"2025111818105543700_bib55","doi-asserted-by":"publisher","first-page":"1025","DOI":"10.1038\/nn.4042","article-title":"A neural network that finds a naturalistic solution for the production of muscle activity","volume":"18","author":"Sussillo","year":"2015","journal-title":"Nature Neuroscience"},{"issue":"24","key":"2025111818105543700_bib56","doi-asserted-by":"publisher","DOI":"10.1063\/1.5011399","article-title":"Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics","volume":"148","author":"Wehmeyer","year":"2018","journal-title":"Journal of Chemical Physics"},{"issue":"1","key":"2025111818105543700_bib57","doi-asserted-by":"publisher","first-page":"25","DOI":"10.1007\/s10107-004-0559-y","article-title":"On the implementation of an interior- point filter line-search algorithm for large-scale nonlinear programming","volume":"106","author":"W\u00e4chter","year":"2006","journal-title":"Mathematical Programming"},{"issue":"1","key":"2025111818105543700_bib58","doi-asserted-by":"publisher","first-page":"016036","DOI":"10.1088\/1741-2552\/ab581a","article-title":"Open source silicon microprobes for high throughput neural recording","volume":"17","author":"Yang","year":"2020","journal-title":"Journal of Neural Engineering"},{"issue":"3","key":"2025111818105543700_bib59","doi-asserted-by":"publisher","first-page":"036013","DOI":"10.1088\/1741-2552\/abcefd","article-title":"Adaptive latent state modeling of brain network dynamics with real-time learning rate optimization","volume":"18","author":"Yang","year":"2021","journal-title":"Journal of Neural Engineering"},{"key":"2025111818105543700_bib60","doi-asserted-by":"publisher","DOI":"10.21203\/rs.3.rs-580874\/v2","author":"Yue","year":"2022","journal-title":"Non-linear model-based control of neural cell dynamics"},{"issue":"6","key":"2025111818105543700_bib61","doi-asserted-by":"publisher","first-page":"2804","DOI":"10.1021\/acs.iecr.2c03691","article-title":"Physics-informed online machine learning and predictive control of nonlinear processes with parameter uncertainty","volume":"62","author":"Zheng","year":"2023","journal-title":"Industrial and Engineering Chemistry Research"}],"container-title":["Neural Computation"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/linproxy.fan.workers.dev:443\/https\/direct.mit.edu\/neco\/article-pdf\/37\/12\/2125\/2560843\/neco.a.37.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/linproxy.fan.workers.dev:443\/https\/direct.mit.edu\/neco\/article-pdf\/37\/12\/2125\/2560843\/neco.a.37.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T23:11:03Z","timestamp":1763507463000},"score":1,"resource":{"primary":{"URL":"https:\/\/linproxy.fan.workers.dev:443\/https\/direct.mit.edu\/neco\/article\/37\/12\/2125\/133751\/Model-Predictive-Control-on-the-Neural-Manifold"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,18]]},"references-count":61,"journal-issue":{"issue":"12","published-online":{"date-parts":[[2025,11,18]]},"published-print":{"date-parts":[[2025,11,18]]}},"URL":"https:\/\/linproxy.fan.workers.dev:443\/https\/doi.org\/10.1162\/neco.a.37","relation":{},"ISSN":["0899-7667","1530-888X"],"issn-type":[{"type":"print","value":"0899-7667"},{"type":"electronic","value":"1530-888X"}],"subject":[],"published-other":{"date-parts":[[2025,12]]},"published":{"date-parts":[[2025,11,18]]}}}