{"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-hardware-in-the-loop-training-a","title":"Neuromorphic Hardware In The Loop: Training a Deep Spiking Network on the BrainScaleS Wafer-Scale System","arxiv_id":"1703.01909","date":"2017-03-06","proceeding":null,"authors":["Sebastian Schmitt","Johann Klaehn","Guillaume Bellec","Andreas Gruebl","Maurice Guettler","Andreas Hartel","Stephan Hartmann","Dan Husmann","Kai Husmann","Vitali Karasenko","Mitja Kleider","Christoph Koke","Christian Mauch","Eric Mueller","Paul Mueller","Johannes Partzsch","Mihai A. Petrovici","Stefan Schiefer","Stefan Scholze","Bernhard Vogginger","Robert Legenstein","Wolfgang Maass","Christian Mayr","Johannes Schemmel","Karlheinz Meier"],"abstract":"Emulating spiking neural networks on analog neuromorphic hardware offers\nseveral advantages over simulating them on conventional computers, particularly\nin terms of speed and energy consumption. However, this usually comes at the\ncost of reduced control over the dynamics of the emulated networks. In this\npaper, we demonstrate how iterative training of a hardware-emulated network can\ncompensate for anomalies induced by the analog substrate. We first convert a\ndeep neural network trained in software to a spiking network on the BrainScaleS\nwafer-scale neuromorphic system, thereby enabling an acceleration factor of 10\n000 compared to the biological time domain. This mapping is followed by the\nin-the-loop training, where in each training step, the network activity is\nfirst recorded in hardware and then used to compute the parameter updates in\nsoftware via backpropagation. An essential finding is that the parameter\nupdates do not have to be precise, but only need to approximately follow the\ncorrect gradient, which simplifies the computation of updates. Using this\napproach, after only several tens of iterations, the spiking network shows an\naccuracy close to the ideal software-emulated prototype. The presented\ntechniques show that deep spiking networks emulated on analog neuromorphic\ndevices can attain good computational performance despite the inherent\nvariations of the analog substrate.","url_abs":"http://arxiv.org/abs/1703.01909v1","url_pdf":"http://arxiv.org/pdf/1703.01909v1.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-hardware-in-the-loop-training-a","repo_url":"https://github.com/hbp-unibi/BS2Cypress","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}