{"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/deep-interest-network-for-click-through-rate","title":"Deep Interest Network for Click-Through Rate Prediction","arxiv_id":"1706.06978","date":"2017-06-21","proceeding":null,"authors":["Guorui Zhou","Chengru Song","Xiaoqiang Zhu","Ying Fan","Han Zhu","Xiao Ma","Yanghui Yan","Junqi Jin","Han Li","Kun Gai"],"abstract":"Click-through rate prediction is an essential task in industrial\napplications, such as online advertising. Recently deep learning based models\nhave been proposed, which follow a similar Embedding\\&MLP paradigm. In these\nmethods large scale sparse input features are first mapped into low dimensional\nembedding vectors, and then transformed into fixed-length vectors in a\ngroup-wise manner, finally concatenated together to fed into a multilayer\nperceptron (MLP) to learn the nonlinear relations among features. In this way,\nuser features are compressed into a fixed-length representation vector, in\nregardless of what candidate ads are. The use of fixed-length vector will be a\nbottleneck, which brings difficulty for Embedding\\&MLP methods to capture\nuser's diverse interests effectively from rich historical behaviors. In this\npaper, we propose a novel model: Deep Interest Network (DIN) which tackles this\nchallenge by designing a local activation unit to adaptively learn the\nrepresentation of user interests from historical behaviors with respect to a\ncertain ad. This representation vector varies over different ads, improving the\nexpressive ability of model greatly. Besides, we develop two techniques:\nmini-batch aware regularization and data adaptive activation function which can\nhelp training industrial deep networks with hundreds of millions of parameters.\nExperiments on two public datasets as well as an Alibaba real production\ndataset with over 2 billion samples demonstrate the effectiveness of proposed\napproaches, which achieve superior performance compared with state-of-the-art\nmethods. DIN now has been successfully deployed in the online display\nadvertising system in Alibaba, serving the main traffic.","url_abs":"http://arxiv.org/abs/1706.06978v4","url_pdf":"http://arxiv.org/pdf/1706.06978v4.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":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/zhougr1993/DeepInterestNetwork","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/GitHub-HongweiZhang/prediction-flow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/StephenBo-China/DIEN-DIN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/StephenBo-China/recommendation_system_sort_model","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/UlionTse/mlgb","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/YafeiWu/DIEN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/YunjiaXi/Multi-Level-Interaction-Reranking","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/imvishvaraj/ctr_nlp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/johnlevi/recsys","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/massquantity/LibRecommender","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/mikudehuane/Configurable-Tensorflow-DIN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/searchlink/din","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/shenweichen/DeepCTR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/shenweichen/DeepCTR-Torch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/tangxyw/RecAlgorithm","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"BSD-2-Clause"}},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/PaddlePaddle/PaddleRec/tree/master/models/rank/din","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/alibaba/TorchEasyRec","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"deep-interest-network-for-click-through-rate","repo_url":"https://github.com/reczoo/FuxiCTR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"click-through-rate-prediction","task_name":"Click-Through Rate Prediction"},{"task_slug":"prediction","task_name":"Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/click-through-rate-prediction-on-amazon","task":"Click-Through Rate Prediction","dataset":"Amazon","model":"DIN + Dice Activation","rank_in_archive_order":1,"of":5,"metrics":{"AUC":"0.8871"},"uses_additional_data":false},{"leaderboard":"/sota/click-through-rate-prediction-on-amazon","task":"Click-Through Rate Prediction","dataset":"Amazon","model":"DIN","rank_in_archive_order":2,"of":5,"metrics":{"AUC":"0.8818"},"uses_additional_data":false},{"leaderboard":"/sota/click-through-rate-prediction-on-movielens","task":"Click-Through Rate Prediction","dataset":"MovieLens 20M","model":"DIN + Dice Activation","rank_in_archive_order":2,"of":6,"metrics":{"AUC":"0.7348"},"uses_additional_data":false},{"leaderboard":"/sota/click-through-rate-prediction-on-movielens","task":"Click-Through Rate Prediction","dataset":"MovieLens 20M","model":"DIN","rank_in_archive_order":3,"of":6,"metrics":{"AUC":"0.7337"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1706.06978","atlas_url":"https://app.syntology.ai/?focus=1706.06978","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}