{"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/skipnet-learning-dynamic-routing-in","title":"SkipNet: Learning Dynamic Routing in Convolutional Networks","arxiv_id":"1711.09485","date":"2017-11-26","proceeding":"ECCV 2018 9","authors":["Xin Wang","Fisher Yu","Zi-Yi Dou","Trevor Darrell","Joseph E. Gonzalez"],"abstract":"While deeper convolutional networks are needed to achieve maximum accuracy in\nvisual perception tasks, for many inputs shallower networks are sufficient. We\nexploit this observation by learning to skip convolutional layers on a\nper-input basis. We introduce SkipNet, a modified residual network, that uses a\ngating network to selectively skip convolutional blocks based on the\nactivations of the previous layer. We formulate the dynamic skipping problem in\nthe context of sequential decision making and propose a hybrid learning\nalgorithm that combines supervised learning and reinforcement learning to\naddress the challenges of non-differentiable skipping decisions. We show\nSkipNet reduces computation by 30-90% while preserving the accuracy of the\noriginal model on four benchmark datasets and outperforms the state-of-the-art\ndynamic networks and static compression methods. We also qualitatively evaluate\nthe gating policy to reveal a relationship between image scale and saliency and\nthe number of layers skipped.","url_abs":"http://arxiv.org/abs/1711.09485v2","url_pdf":"http://arxiv.org/pdf/1711.09485v2.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":"skipnet-learning-dynamic-routing-in","repo_url":"https://github.com/ucbdrive/skipnet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"skipnet-learning-dynamic-routing-in","repo_url":"https://github.com/geekJZY/arcticnet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"decision-making","task_name":"Decision Making"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"sequential-decision-making","task_name":"Sequential Decision Making"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1711.09485","atlas_url":"https://app.syntology.ai/?focus=1711.09485","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}