{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/dendritic-error-backpropagation-in-deep","title":"Dendritic error backpropagation in deep cortical microcircuits","arxiv_id":"1801.00062","date":"2017-12-30","proceeding":null,"authors":["João Sacramento","Rui Ponte Costa","Yoshua Bengio","Walter Senn"],"abstract":"Animal behaviour depends on learning to associate sensory stimuli with the\ndesired motor command. Understanding how the brain orchestrates the necessary\nsynaptic modifications across different brain areas has remained a longstanding\npuzzle. Here, we introduce a multi-area neuronal network model in which\nsynaptic plasticity continuously adapts the network towards a global desired\noutput. In this model synaptic learning is driven by a local dendritic\nprediction error that arises from a failure to predict the top-down input given\nthe bottom-up activities. Such errors occur at apical dendrites of pyramidal\nneurons where both long-range excitatory feedback and local inhibitory\npredictions are integrated. When local inhibition fails to match excitatory\nfeedback an error occurs which triggers plasticity at bottom-up synapses at\nbasal dendrites of the same pyramidal neurons. We demonstrate the learning\ncapabilities of the model in a number of tasks and show that it approximates\nthe classical error backpropagation algorithm. Finally, complementing this\ncortical circuit with a disinhibitory mechanism enables attention-like stimulus\ndenoising and generation. Our framework makes several experimental predictions\non the function of dendritic integration and cortical microcircuits, is\nconsistent with recent observations of cross-area learning, and suggests a\nbiological implementation of deep learning.","url_abs":"http://arxiv.org/abs/1801.00062v1","url_pdf":"http://arxiv.org/pdf/1801.00062v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"dendritic-error-backpropagation-in-deep","repo_url":"https://github.com/EntropicEffect/dendritic_backprop","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"denoising","task_name":"Denoising"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1801.00062","atlas_url":"https://app.syntology.ai/?focus=1801.00062","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}