{"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/atrank-an-attention-based-user-behavior","title":"ATRank: An Attention-Based User Behavior Modeling Framework for Recommendation","arxiv_id":"1711.06632","date":"2017-11-17","proceeding":null,"authors":["Chang Zhou","Jinze Bai","Junshuai Song","Xiaofei Liu","Zhengchao Zhao","Xiusi Chen","Jun Gao"],"abstract":"A user can be represented as what he/she does along the history. A common way\nto deal with the user modeling problem is to manually extract all kinds of\naggregated features over the heterogeneous behaviors, which may fail to fully\nrepresent the data itself due to limited human instinct. Recent works usually\nuse RNN-based methods to give an overall embedding of a behavior sequence,\nwhich then could be exploited by the downstream applications. However, this can\nonly preserve very limited information, or aggregated memories of a person.\nWhen a downstream application requires to facilitate the modeled user features,\nit may lose the integrity of the specific highly correlated behavior of the\nuser, and introduce noises derived from unrelated behaviors. This paper\nproposes an attention based user behavior modeling framework called ATRank,\nwhich we mainly use for recommendation tasks. Heterogeneous user behaviors are\nconsidered in our model that we project all types of behaviors into multiple\nlatent semantic spaces, where influence can be made among the behaviors via\nself-attention. Downstream applications then can use the user behavior vectors\nvia vanilla attention. Experiments show that ATRank can achieve better\nperformance and faster training process. We further explore ATRank to use one\nunified model to predict different types of user behaviors at the same time,\nshowing a comparable performance with the highly optimized individual models.","url_abs":"http://arxiv.org/abs/1711.06632v2","url_pdf":"http://arxiv.org/pdf/1711.06632v2.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":"atrank-an-attention-based-user-behavior","repo_url":"https://github.com/jinze1994/ATRank","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"atrank-an-attention-based-user-behavior","repo_url":"https://github.com/johnlevi/recsys","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1711.06632","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}