{"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/deepcode-feedback-codes-via-deep-learning","title":"Deepcode: Feedback Codes via Deep Learning","arxiv_id":"1807.00801","date":"2018-07-02","proceeding":"NeurIPS 2018 12","authors":["Hyeji Kim","Yihan Jiang","Sreeram Kannan","Sewoong Oh","Pramod Viswanath"],"abstract":"The design of codes for communicating reliably over a statistically well\ndefined channel is an important endeavor involving deep mathematical research\nand wide-ranging practical applications. In this work, we present the first\nfamily of codes obtained via deep learning, which significantly beats\nstate-of-the-art codes designed over several decades of research. The\ncommunication channel under consideration is the Gaussian noise channel with\nfeedback, whose study was initiated by Shannon; feedback is known theoretically\nto improve reliability of communication, but no practical codes that do so have\never been successfully constructed.\n  We break this logjam by integrating information theoretic insights\nharmoniously with recurrent-neural-network based encoders and decoders to\ncreate novel codes that outperform known codes by 3 orders of magnitude in\nreliability. We also demonstrate several desirable properties of the codes: (a)\ngeneralization to larger block lengths, (b) composability with known codes, (c)\nadaptation to practical constraints. This result also has broader ramifications\nfor coding theory: even when the channel has a clear mathematical model, deep\nlearning methodologies, when combined with channel-specific\ninformation-theoretic insights, can potentially beat state-of-the-art codes\nconstructed over decades of mathematical research.","url_abs":"http://arxiv.org/abs/1807.00801v1","url_pdf":"http://arxiv.org/pdf/1807.00801v1.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":"deepcode-feedback-codes-via-deep-learning","repo_url":"https://github.com/hyejikim1/Deepcode","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.00801","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}