{"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/improving-noise-tolerance-of-mixed-signal","title":"Improving Noise Tolerance of Mixed-Signal Neural Networks","arxiv_id":"1904.01705","date":"2019-04-02","proceeding":null,"authors":["Michael Klachko","Mohammad Reza Mahmoodi","Dmitri B. Strukov"],"abstract":"Mixed-signal hardware accelerators for deep learning achieve orders of\nmagnitude better power efficiency than their digital counterparts. In the\nultra-low power consumption regime, limited signal precision inherent to analog\ncomputation becomes a challenge. We perform a case study of a 6-layer\nconvolutional neural network running on a mixed-signal accelerator and evaluate\nits sensitivity to hardware specific noise. We apply various methods to improve\nnoise robustness of the network and demonstrate an effective way to optimize\nuseful signal ranges through adaptive signal clipping. The resulting model is\nrobust enough to achieve 80.2% classification accuracy on CIFAR-10 dataset with\njust 1.4 mW power budget, while 6 mW budget allows us to achieve 87.1%\naccuracy, which is within 1% of the software baseline. For comparison, the\nunoptimized version of the same model achieves only 67.7% accuracy at 1.4 mW\nand 78.6% at 6 mW.","url_abs":"http://arxiv.org/abs/1904.01705v1","url_pdf":"http://arxiv.org/pdf/1904.01705v1.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":"improving-noise-tolerance-of-mixed-signal","repo_url":"https://github.com/michaelklachko/noisynet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}