{"database":"biostudies-literature","file_versions":[],"scores":null,"additional":{"submitter":["Aitchison L"],"funding":["Swiss National Science Foundation","Simons Collaboration for the Global Brain","UCL Graduate Research and UCL Overseas Research Scholarships","Wellcome Trust","Gatsby Charitable Foundation"],"pagination":["565-571"],"full_dataset_link":["https://www.ebi.ac.uk/biostudies/studies/S-EPMC7617048"],"repository":["biostudies-literature"],"omics_type":["Unknown"],"volume":["24(4)"],"pubmed_abstract":["Learning, especially rapid learning, is critical for survival. However, learning is hard; a large number of synaptic weights must be set based on noisy, often ambiguous, sensory information. In such a high-noise regime, keeping track of probability distributions over weights is the optimal strategy. Here we hypothesize that synapses take that strategy; in essence, when they estimate weights, they include error bars. They then use that uncertainty to adjust their learning rates, with more uncertain weights having higher learning rates. We also make a second, independent, hypothesis: synapses communicate their uncertainty by linking it to variability in postsynaptic potential size, with more uncertainty leading to more variability. These two hypotheses cast synaptic plasticity as a problem of Bayesian inference, and thus provide a normative view of learning. They generalize known learning rules, offer an explanation for the large variability in the size of postsynaptic potentials and make falsifiable experimental predictions."],"journal":["Nature neuroscience"],"pubmed_title":["Synaptic plasticity as Bayesian inference."],"pmcid":["PMC7617048"],"funding_grant_id":["150637","110114","PP00P3 150637","110114/Z/15/Z"],"pubmed_authors":["Jegminat J","Pfister JP","Pouget A","Aitchison L","Menendez JA","Latham PE"],"additional_accession":[]},"is_claimable":false,"name":"Synaptic plasticity as Bayesian inference.","description":"Learning, especially rapid learning, is critical for survival. However, learning is hard; a large number of synaptic weights must be set based on noisy, often ambiguous, sensory information. In such a high-noise regime, keeping track of probability distributions over weights is the optimal strategy. Here we hypothesize that synapses take that strategy; in essence, when they estimate weights, they include error bars. They then use that uncertainty to adjust their learning rates, with more uncertain weights having higher learning rates. We also make a second, independent, hypothesis: synapses communicate their uncertainty by linking it to variability in postsynaptic potential size, with more uncertainty leading to more variability. These two hypotheses cast synaptic plasticity as a problem of Bayesian inference, and thus provide a normative view of learning. They generalize known learning rules, offer an explanation for the large variability in the size of postsynaptic potentials and make falsifiable experimental predictions.","dates":{"release":"2021-01-01T00:00:00Z","publication":"2021 Apr","modification":"2025-04-19T17:28:05.258Z","creation":"2025-04-19T17:28:05.258Z"},"accession":"S-EPMC7617048","cross_references":{"pubmed":["33707754"],"doi":["10.1038/s41593-021-00809-5"]}}