{"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/depthwise-separable-convolutions-for-neural","title":"Depthwise Separable Convolutions for Neural Machine Translation","arxiv_id":"1706.03059","date":"2017-06-09","proceeding":"ICLR 2018 1","authors":["Lukasz Kaiser","Aidan N. Gomez","Francois Chollet"],"abstract":"Depthwise separable convolutions reduce the number of parameters and\ncomputation used in convolutional operations while increasing representational\nefficiency. They have been shown to be successful in image classification\nmodels, both in obtaining better models than previously possible for a given\nparameter count (the Xception architecture) and considerably reducing the\nnumber of parameters required to perform at a given level (the MobileNets\nfamily of architectures). Recently, convolutional sequence-to-sequence networks\nhave been applied to machine translation tasks with good results. In this work,\nwe study how depthwise separable convolutions can be applied to neural machine\ntranslation. We introduce a new architecture inspired by Xception and ByteNet,\ncalled SliceNet, which enables a significant reduction of the parameter count\nand amount of computation needed to obtain results like ByteNet, and, with a\nsimilar parameter count, achieves new state-of-the-art results. In addition to\nshowing that depthwise separable convolutions perform well for machine\ntranslation, we investigate the architectural changes that they enable: we\nobserve that thanks to depthwise separability, we can increase the length of\nconvolution windows, removing the need for filter dilation. We also introduce a\nnew \"super-separable\" convolution operation that further reduces the number of\nparameters and computational cost for obtaining state-of-the-art results.","url_abs":"http://arxiv.org/abs/1706.03059v2","url_pdf":"http://arxiv.org/pdf/1706.03059v2.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":"depthwise-separable-convolutions-for-neural","repo_url":"https://github.com/tensorflow/tensor2tensor","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"depthwise-separable-convolutions-for-neural","repo_url":"https://github.com/outcastofmusic/quick-nlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/machine-translation-on-wmt2014-english-german","task":"Machine Translation","dataset":"WMT2014 English-German","model":"SliceNet","rank_in_archive_order":62,"of":91,"metrics":{"BLEU score":"26.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.03059","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}