{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,15]],"date-time":"2025-08-15T01:31:32Z","timestamp":1755221492543,"version":"3.43.0"},"reference-count":36,"publisher":"MIT Press","issue":"9","content-domain":{"domain":["direct.mit.edu"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2025,8,8]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Understanding how brain networks learn and manage multiple tasks simultaneously is of interest in both neuroscience and artificial intelligence. In this regard, a recent research thread in theoretical neuroscience has focused on how recurrent neural network models and their internal dynamics enact multitask learning. To manage different tasks requires a mechanism to convey information about task identity or context into the model, which from a biological perspective may involve mechanisms of neuromodulation. In this study, we use recurrent network models to probe the distinctions between two forms of contextual modulation of neural dynamics, at the level of neuronal excitability and at the level of synaptic strength. We characterize these mechanisms in terms of their functional outcomes, focusing on their robustness to context ambiguity and, relatedly, their efficiency with respect to packing multiple tasks into finite-size networks. We also demonstrate the distinction between these mechanisms at the level of the neuronal dynamics they induce. Together, these characterizations indicate complementarity and synergy in how these mechanisms act, potentially over many timescales, toward enhancing the robustness of multitask learning.<\/jats:p>","DOI":"10.1162\/neco.a.18","type":"journal-article","created":{"date-parts":[[2025,7,24]],"date-time":"2025-07-24T15:04:51Z","timestamp":1753369491000},"page":"1529-1550","update-policy":"https:\/\/doi.org\/10.1162\/mitpressjournals.corrections.policy","source":"Crossref","is-referenced-by-count":0,"title":["Synergistic Pathways of Modulation Enable Robust Task Packing Within Neural Dynamics"],"prefix":"10.1162","volume":"37","author":[{"given":"Giacomo","family":"Vedovati","sequence":"first","affiliation":[{"name":"Department of Electrical and Systems Engineering, Washington University in St. Louis, St. Louis, MO 63130, USA g.vedovati@wustl.edu"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"ShiNung","family":"Ching","sequence":"additional","affiliation":[{"name":"Department of Electrical and Systems Engineering, Washington University in St. Louis, St. Louis, MO 63130, USA shinung@wustl.edu"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"281","published-online":{"date-parts":[[2025,8,8]]},"reference":[{"key":"2025080816010080800_bib1","doi-asserted-by":"crossref","DOI":"10.7554\/eLife.83035","article-title":"Neural population dynamics of computing with synaptic modulations","volume":"12","author":"Aitken","year":"2023","journal-title":"eLife"},{"key":"2025080816010080800_bib2","article-title":"Using fast weights to attend to the recent past","volume-title":"Advances in neural information processing systems","author":"Ba","year":"2016"},{"issue":"1","key":"2025080816010080800_bib3","doi-asserted-by":"publisher","first-page":"105","DOI":"10.1016\/0959-4388(91)90017-2","article-title":"Ionic conductance mechanisms contributing to the electrophysiological properties of neurons","volume":"1","author":"Baxter","year":"1991","journal-title":"Current Opinion in Neurobiology"},{"key":"2025080816010080800_bib4","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v32i1.11273","article-title":"CA-RNN: Using context-aligned recurrent neural networks for modeling sentence similarity","volume-title":"Proceedings of the AAAI Conference on Artificial Intelligence","author":"Chen","year":"2018"},{"issue":"7","key":"2025080816010080800_bib5","doi-asserted-by":"publisher","first-page":"811","DOI":"10.1038\/nn.2842","article-title":"Measuring and interpreting neuronal correlations","volume":"14","author":"Cohen","year":"2011","journal-title":"Nature Neuroscience"},{"issue":"1","key":"2025080816010080800_bib6","doi-asserted-by":"publisher","first-page":"162","DOI":"10.1016\/j.neuron.2008.08.007","article-title":"Context-dependent changes in functional circuitry in visual area MT","volume":"60","author":"Cohen","year":"2008","journal-title":"Neuron"},{"key":"2025080816010080800_bib7","article-title":"Computational models of attention and cognitive control","volume-title":"Cambridge handbook of computational psychology","author":"De Pisapia","year":"2008"},{"key":"2025080816010080800_bib8","first-page":"2393","article-title":"Context contrasted feature and gated multi-scale aggregation for scene segmentation","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Ding","year":"2018"},{"issue":"7","key":"2025080816010080800_bib9","doi-asserted-by":"publisher","first-page":"1349","DOI":"10.1038\/s41593-024-01668-6","article-title":"Flexible multitask computation in recurrent networks utilizes shared dynamical motifs","volume":"27","author":"Driscoll","year":"2024","journal-title":"Nature Neuroscience"},{"key":"2025080816010080800_bib10","article-title":"Dynamical system approach to explainability in recurrent neural networks","volume-title":"CNIA 2021: Conf\u00e9rence Nationale en Intelligence Artificielle","author":"Dubreuil","year":"2021"},{"issue":"6","key":"2025080816010080800_bib11","doi-asserted-by":"publisher","first-page":"783","DOI":"10.1038\/s41593-022-01088-4","article-title":"The role of population structure in computations through neural dynamics","volume":"25","author":"Dubreuil","year":"2022","journal-title":"Nature Neuroscience"},{"issue":"1","key":"2025080816010080800_bib12","first-page":"723","article-title":"A kernel two-sample test","volume":"13","author":"Gretton","year":"2012","journal-title":"Journal of Machine Learning Research"},{"key":"2025080816010080800_bib13","doi-asserted-by":"publisher","DOI":"10.3389\/fnbeh.2012.00050","article-title":"Context, emotion, and the strategic pursuit of goals: Interactions among multiple brain systems controlling motivated behavior","volume":"6","author":"Gruber","year":"2012","journal-title":"Frontiers in Behavioral Neuroscience"},{"issue":"7278","key":"2025080816010080800_bib14","doi-asserted-by":"publisher","first-page":"232","DOI":"10.1038\/nature08673","article-title":"Long-term potentiation depends on release of D-serine from astrocytes","volume":"463","author":"Henneberger","year":"2010","journal-title":"Nature"},{"issue":"11","key":"2025080816010080800_bib15","doi-asserted-by":"publisher","first-page":"7087","DOI":"10.1523\/JNEUROSCI.14-11-07087.1994","article-title":"Role of neuronal synchronizing mechanisms in the propagation of spreading depression in the in vivo hippocampus","volume":"14","author":"Herreras","year":"1994","journal-title":"Journal of Neuroscience"},{"journal-title":"Gated embeddings in end-to-end speech recognition for conversational-context fusion.","year":"2019","author":"Kim","key":"2025080816010080800_bib16"},{"journal-title":"Adam: A method for stochastic optimization.","year":"2014","author":"Kingma","key":"2025080816010080800_bib17"},{"issue":"1","key":"2025080816010080800_bib18","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevX.12.011011","article-title":"Theory of gating in recurrent neural networks","volume":"12","author":"Krishnamurthy","year":"2022","journal-title":"Physical Review X"},{"issue":"9","key":"2025080816010080800_bib19","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2210622120","article-title":"Bayesian inference in ring attractor networks","volume":"120","author":"Kutschireiter","year":"2023","journal-title":"Proceedings of the National Academy of Sciences"},{"journal-title":"Norepinephrine signals through astrocytes to modulate synapses.","year":"2024","author":"Lefton","key":"2025080816010080800_bib20"},{"issue":"1","key":"2025080816010080800_bib21","doi-asserted-by":"publisher","first-page":"87","DOI":"10.1152\/physrev.00014.2003","article-title":"Long-term potentiation and memory","volume":"84","author":"Lynch","year":"2004","journal-title":"Physiological Reviews"},{"issue":"7474","key":"2025080816010080800_bib22","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":"44","key":"2025080816010080800_bib23","doi-asserted-by":"publisher","first-page":"E10467","DOI":"10.1073\/pnas.1803839115","article-title":"Alleviating catastrophic forgetting using context-dependent gating and synaptic stabilization","volume":"115","author":"Masse","year":"2018","journal-title":"Proceedings of the National Academy of Sciences"},{"issue":"3","key":"2025080816010080800_bib24","doi-asserted-by":"publisher","first-page":"609","DOI":"10.1016\/j.neuron.2018.07.003","article-title":"Linking connectivity, dynamics, and computations in low-rank recurrent neural networks","volume":"99","author":"Mastrogiuseppe","year":"2018","journal-title":"Neuron"},{"issue":"11","key":"2025080816010080800_bib25","doi-asserted-by":"publisher","first-page":"1848","DOI":"10.1038\/s41593-023-01448-8","article-title":"A conceptual framework for astrocyte function","volume":"26","author":"Murphy-Royal","year":"2023","journal-title":"Nature Neuroscience"},{"key":"2025080816010080800_bib26","doi-asserted-by":"publisher","first-page":"48","DOI":"10.1016\/j.conb.2014.05.003","article-title":"Neuromodulation of neurons and synapses","volume":"29","author":"Nadim","year":"2014","journal-title":"Current Opinion in Neurobiology"},{"issue":"4","key":"2025080816010080800_bib27","doi-asserted-by":"publisher","first-page":"576","DOI":"10.1016\/j.neuron.2020.12.008","article-title":"Behaviorally consequential astrocytic regulation of neural circuits","volume":"109","author":"Nagai","year":"2021","journal-title":"Neuron"},{"issue":"1715","key":"2025080816010080800_bib28","doi-asserted-by":"publisher","DOI":"10.1098\/rstb.2016.0154","article-title":"Astrocytic control of synaptic function","volume":"372","author":"Papouin","year":"2017","journal-title":"Philosophical Transactions of the Royal Society B: Biological Sciences"},{"issue":"2","key":"2025080816010080800_bib29","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pcbi.1004792","article-title":"Training excitatory-inhibitory recurrent neural networks for cognitive tasks: A simple and flexible framework","volume":"12","author":"Song","year":"2016","journal-title":"PLOS Computational Biology"},{"issue":"12","key":"2025080816010080800_bib30","doi-asserted-by":"publisher","first-page":"1774","DOI":"10.1038\/s41593-018-0276-0","article-title":"Motor primitives in space and time via targeted gain modulation in cortical networks","volume":"21","author":"Stroud","year":"2018","journal-title":"Nature Neuroscience"},{"journal-title":"Neuromodulators generate multiple context-relevant behaviors in a recurrent neural network by shifting activity flows in hyperchannels","year":"2021","author":"Tsuda","key":"2025080816010080800_bib31"},{"issue":"6","key":"2025080816010080800_bib32","doi-asserted-by":"publisher","DOI":"10.1088\/0143-0807\/37\/6\/065001","article-title":"Visualizing rotations and composition of rotations with the Rodrigues vector","volume":"37","author":"Valdenebro","year":"2016","journal-title":"European Journal of Physics"},{"key":"2025080816010080800_bib33","first-page":"24072","article-title":"Extracting computational mechanisms from neural data using low-rank RNNS","volume-title":"Advances in neural information processing systems","author":"Valente","year":"2022"},{"issue":"1","key":"2025080816010080800_bib34","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0227922","article-title":"Introducing neuromodulation in deep neural networks to learn adaptive behaviours","volume":"15","author":"Vecoven","year":"2020","journal-title":"PLOS One"},{"issue":"2","key":"2025080816010080800_bib35","doi-asserted-by":"publisher","first-page":"297","DOI":"10.1038\/s41593-018-0310-2","article-title":"Task representations in neural networks trained to perform many cognitive tasks","volume":"22","author":"Yang","year":"2019","journal-title":"Nature Neuroscience"},{"issue":"1","key":"2025080816010080800_bib36","article-title":"Firing feature-driven neural circuits with scalable memristive neurons for robotic obstacle avoidance","volume":"15","author":"Yang","year":"2024","journal-title":"Nature Communications"}],"container-title":["Neural Computation"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/37\/9\/1529\/2538525\/neco.a.18.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/direct.mit.edu\/neco\/article-pdf\/37\/9\/1529\/2538525\/neco.a.18.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,8]],"date-time":"2025-08-08T20:01:10Z","timestamp":1754683270000},"score":1,"resource":{"primary":{"URL":"https:\/\/direct.mit.edu\/neco\/article\/37\/9\/1529\/131937\/Synergistic-Pathways-of-Modulation-Enable-Robust"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,8,8]]},"references-count":36,"journal-issue":{"issue":"9","published-online":{"date-parts":[[2025,8,8]]},"published-print":{"date-parts":[[2025,8,8]]}},"URL":"https:\/\/doi.org\/10.1162\/neco.a.18","relation":{},"ISSN":["0899-7667","1530-888X"],"issn-type":[{"type":"print","value":"0899-7667"},{"type":"electronic","value":"1530-888X"}],"subject":[],"published-other":{"date-parts":[[2025,9]]},"published":{"date-parts":[[2025,8,8]]}}}