{"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-and-inference-with-integers-in-deep","title":"Training and Inference with Integers in Deep Neural Networks","arxiv_id":"1802.04680","date":"2018-02-13","proceeding":"ICLR 2018 1","authors":["Shuang Wu","Guoqi Li","Feng Chen","Luping Shi"],"abstract":"Researches on deep neural networks with discrete parameters and their\ndeployment in embedded systems have been active and promising topics. Although\nprevious works have successfully reduced precision in inference, transferring\nboth training and inference processes to low-bitwidth integers has not been\ndemonstrated simultaneously. In this work, we develop a new method termed as\n\"WAGE\" to discretize both training and inference, where weights (W),\nactivations (A), gradients (G) and errors (E) among layers are shifted and\nlinearly constrained to low-bitwidth integers. To perform pure discrete\ndataflow for fixed-point devices, we further replace batch normalization by a\nconstant scaling layer and simplify other components that are arduous for\ninteger implementation. Improved accuracies can be obtained on multiple\ndatasets, which indicates that WAGE somehow acts as a type of regularization.\nEmpirically, we demonstrate the potential to deploy training in hardware\nsystems such as integer-based deep learning accelerators and neuromorphic chips\nwith comparable accuracy and higher energy efficiency, which is crucial to\nfuture AI applications in variable scenarios with transfer and continual\nlearning demands.","url_abs":"http://arxiv.org/abs/1802.04680v1","url_pdf":"http://arxiv.org/pdf/1802.04680v1.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-and-inference-with-integers-in-deep","repo_url":"https://github.com/boluoweifenda/WAGE","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"training-and-inference-with-integers-in-deep","repo_url":"https://github.com/Tiiiger/QPyTorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"training-and-inference-with-integers-in-deep","repo_url":"https://github.com/stevenygd/WAGE.pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"continual-learning","task_name":"Continual Learning"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.04680","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}