{"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/neuromorphic-deep-learning-machines","title":"Neuromorphic Deep Learning Machines","arxiv_id":"1612.05596","date":"2016-12-16","proceeding":null,"authors":["Emre Neftci","Charles Augustine","Somnath Paul","Georgios Detorakis"],"abstract":"An ongoing challenge in neuromorphic computing is to devise general and\ncomputationally efficient models of inference and learning which are compatible\nwith the spatial and temporal constraints of the brain. One increasingly\npopular and successful approach is to take inspiration from inference and\nlearning algorithms used in deep neural networks. However, the workhorse of\ndeep learning, the gradient descent Back Propagation (BP) rule, often relies on\nthe immediate availability of network-wide information stored with\nhigh-precision memory, and precise operations that are difficult to realize in\nneuromorphic hardware. Remarkably, recent work showed that exact backpropagated\nweights are not essential for learning deep representations. Random BP replaces\nfeedback weights with random ones and encourages the network to adjust its\nfeed-forward weights to learn pseudo-inverses of the (random) feedback weights.\nBuilding on these results, we demonstrate an event-driven random BP (eRBP) rule\nthat uses an error-modulated synaptic plasticity for learning deep\nrepresentations in neuromorphic computing hardware. The rule requires only one\naddition and two comparisons for each synaptic weight using a two-compartment\nleaky Integrate & Fire (I&F) neuron, making it very suitable for implementation\nin digital or mixed-signal neuromorphic hardware. Our results show that using\neRBP, deep representations are rapidly learned, achieving nearly identical\nclassification accuracies compared to artificial neural network simulations on\nGPUs, while being robust to neural and synaptic state quantizations during\nlearning.","url_abs":"http://arxiv.org/abs/1612.05596v2","url_pdf":"http://arxiv.org/pdf/1612.05596v2.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":"neuromorphic-deep-learning-machines","repo_url":"https://gitlab.com/eneftci/erbp_auryn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1612.05596","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}