{"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/decoupled-networks","title":"Decoupled Networks","arxiv_id":"1804.08071","date":"2018-04-22","proceeding":"CVPR 2018 6","authors":["Weiyang Liu","Zhen Liu","Zhiding Yu","Bo Dai","Rongmei Lin","Yisen Wang","James M. Rehg","Le Song"],"abstract":"Inner product-based convolution has been a central component of convolutional\nneural networks (CNNs) and the key to learning visual representations. Inspired\nby the observation that CNN-learned features are naturally decoupled with the\nnorm of features corresponding to the intra-class variation and the angle\ncorresponding to the semantic difference, we propose a generic decoupled\nlearning framework which models the intra-class variation and semantic\ndifference independently. Specifically, we first reparametrize the inner\nproduct to a decoupled form and then generalize it to the decoupled convolution\noperator which serves as the building block of our decoupled networks. We\npresent several effective instances of the decoupled convolution operator. Each\ndecoupled operator is well motivated and has an intuitive geometric\ninterpretation. Based on these decoupled operators, we further propose to\ndirectly learn the operator from data. Extensive experiments show that such\ndecoupled reparameterization renders significant performance gain with easier\nconvergence and stronger robustness.","url_abs":"http://arxiv.org/abs/1804.08071v1","url_pdf":"http://arxiv.org/pdf/1804.08071v1.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":"decoupled-networks","repo_url":"https://github.com/yujiacheng333/BaseDcLayer","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.08071","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}