{"department":[{"_id":"TiVo"}],"main_file_link":[{"url":"https://proceedings.neurips.cc/paper/2020/hash/bdbd5ebfde4934142c8a88e7a3796cd5-Abstract.html","open_access":"1"}],"abstract":[{"text":"The search for biologically faithful synaptic plasticity rules has resulted in a large body of models. They are usually inspired by – and fitted to – experimental data, but they rarely produce neural dynamics that serve complex functions. These failures suggest that current plasticity models are still under-constrained by existing data. Here, we present an alternative approach that uses meta-learning to discover plausible synaptic plasticity rules. Instead of experimental data, the rules are constrained by the functions they implement and the structure they are meant to produce. Briefly, we parameterize synaptic plasticity rules by a Volterra expansion and then use supervised learning methods (gradient descent or evolutionary strategies) to minimize a problem-dependent loss function that quantifies how effectively a candidate plasticity rule transforms an initially random network into one with the desired function. We first validate our approach by re-discovering previously described plasticity rules, starting at the single-neuron level and “Oja’s rule”, a simple Hebbian plasticity rule that captures the direction of most variability of inputs to a neuron (i.e., the first principal component). We expand the problem to the network level and ask the framework to find Oja’s rule together with an anti-Hebbian rule such that an initially random two-layer firing-rate network will recover several principal components of the input space after learning. Next, we move to networks of integrate-and-fire neurons with plastic inhibitory afferents. We train for rules that achieve a target firing rate by countering tuned excitation. Our algorithm discovers a specific subset of the manifold of rules that can solve this task. Our work is a proof of principle of an automated and unbiased approach to unveil synaptic plasticity rules that obey biological constraints and can solve complex functions.","lang":"eng"}],"publication_status":"published","date_published":"2020-12-06T00:00:00Z","article_processing_charge":"No","month":"12","date_updated":"2023-10-18T09:20:55Z","type":"conference","day":"06","title":"A meta-learning approach to (re)discover plasticity rules that carve a desired function into a neural network","quality_controlled":"1","related_material":{"link":[{"relation":"is_continued_by","url":"https://doi.org/10.1101/2020.10.24.353409"}],"record":[{"status":"public","relation":"dissertation_contains","id":"14422"}]},"status":"public","scopus_import":"1","oa":1,"oa_version":"Published Version","user_id":"6785fbc1-c503-11eb-8a32-93094b40e1cf","date_created":"2021-07-04T22:01:27Z","language":[{"iso":"eng"}],"_id":"9633","ec_funded":1,"conference":{"end_date":"2020-12-12","location":"Vancouver, Canada","name":"NeurIPS: Conference on Neural Information Processing Systems","start_date":"2020-12-06"},"citation":{"ama":"Confavreux BJ, Zenke F, Agnes EJ, Lillicrap T, Vogels TP. A meta-learning approach to (re)discover plasticity rules that carve a desired function into a neural network. In: <i>Advances in Neural Information Processing Systems</i>. Vol 33. ; 2020:16398-16408.","apa":"Confavreux, B. J., Zenke, F., Agnes, E. J., Lillicrap, T., &#38; Vogels, T. P. (2020). A meta-learning approach to (re)discover plasticity rules that carve a desired function into a neural network. In <i>Advances in Neural Information Processing Systems</i> (Vol. 33, pp. 16398–16408). Vancouver, Canada.","ista":"Confavreux BJ, Zenke F, Agnes EJ, Lillicrap T, Vogels TP. 2020. A meta-learning approach to (re)discover plasticity rules that carve a desired function into a neural network. Advances in Neural Information Processing Systems. NeurIPS: Conference on Neural Information Processing Systems vol. 33, 16398–16408.","chicago":"Confavreux, Basile J, Friedemann Zenke, Everton J. Agnes, Timothy Lillicrap, and Tim P Vogels. “A Meta-Learning Approach to (Re)Discover Plasticity Rules That Carve a Desired Function into a Neural Network.” In <i>Advances in Neural Information Processing Systems</i>, 33:16398–408, 2020.","short":"B.J. Confavreux, F. Zenke, E.J. Agnes, T. Lillicrap, T.P. Vogels, in:, Advances in Neural Information Processing Systems, 2020, pp. 16398–16408.","mla":"Confavreux, Basile J., et al. “A Meta-Learning Approach to (Re)Discover Plasticity Rules That Carve a Desired Function into a Neural Network.” <i>Advances in Neural Information Processing Systems</i>, vol. 33, 2020, pp. 16398–408.","ieee":"B. J. Confavreux, F. Zenke, E. J. Agnes, T. Lillicrap, and T. P. Vogels, “A meta-learning approach to (re)discover plasticity rules that carve a desired function into a neural network,” in <i>Advances in Neural Information Processing Systems</i>, Vancouver, Canada, 2020, vol. 33, pp. 16398–16408."},"publication_identifier":{"issn":["1049-5258"]},"page":"16398-16408","publication":"Advances in Neural Information Processing Systems","author":[{"last_name":"Confavreux","first_name":"Basile J","id":"C7610134-B532-11EA-BD9F-F5753DDC885E","full_name":"Confavreux, Basile J"},{"last_name":"Zenke","first_name":"Friedemann","full_name":"Zenke, Friedemann"},{"full_name":"Agnes, Everton J.","last_name":"Agnes","first_name":"Everton J."},{"full_name":"Lillicrap, Timothy","last_name":"Lillicrap","first_name":"Timothy"},{"first_name":"Tim P","last_name":"Vogels","full_name":"Vogels, Tim P","id":"CB6FF8D2-008F-11EA-8E08-2637E6697425","orcid":"0000-0003-3295-6181"}],"project":[{"grant_number":"214316/Z/18/Z","_id":"c084a126-5a5b-11eb-8a69-d75314a70a87","name":"What’s in a memory? Spatiotemporal dynamics in strongly coupled recurrent neuronal networks."},{"call_identifier":"H2020","grant_number":"819603","name":"Learning the shape of synaptic plasticity rules for neuronal architectures and function through machine learning.","_id":"0aacfa84-070f-11eb-9043-d7eb2c709234"}],"acknowledgement":"We would like to thank Chaitanya Chintaluri, Georgia Christodoulou, Bill Podlaski and Merima Šabanovic for useful discussions and comments. This work was supported by a Wellcome Trust ´ Senior Research Fellowship (214316/Z/18/Z), a BBSRC grant (BB/N019512/1), an ERC consolidator Grant (SYNAPSEEK), a Leverhulme Trust Project Grant (RPG-2016-446), and funding from École Polytechnique, Paris.","year":"2020","intvolume":"        33","volume":33}