{"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/training-deep-neural-networks-with-low","title":"Training deep neural networks with low precision multiplications","arxiv_id":"1412.7024","date":"2014-12-22","proceeding":null,"authors":["Matthieu Courbariaux","Yoshua Bengio","Jean-Pierre David"],"abstract":"Multipliers are the most space and power-hungry arithmetic operators of the\ndigital implementation of deep neural networks. We train a set of\nstate-of-the-art neural networks (Maxout networks) on three benchmark datasets:\nMNIST, CIFAR-10 and SVHN. They are trained with three distinct formats:\nfloating point, fixed point and dynamic fixed point. For each of those datasets\nand for each of those formats, we assess the impact of the precision of the\nmultiplications on the final error after training. We find that very low\nprecision is sufficient not just for running trained networks but also for\ntraining them. For example, it is possible to train Maxout networks with 10\nbits multiplications.","url_abs":"http://arxiv.org/abs/1412.7024v5","url_pdf":"http://arxiv.org/pdf/1412.7024v5.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":"training-deep-neural-networks-with-low","repo_url":"https://github.com/MatthieuCourbariaux/deep-learning-multipliers","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"GPL-2.0"}}],"tasks":[],"methods":[{"method_slug":"maxout","method_name":"Maxout"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1412.7024","atlas_url":"https://app.syntology.ai/?focus=1412.7024","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}