{"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/neural-factorization-machines-for-sparse","title":"Neural Factorization Machines for Sparse Predictive Analytics","arxiv_id":"1708.05027","date":"2017-08-16","proceeding":null,"authors":["Xiangnan He","Tat-Seng Chua"],"abstract":"Many predictive tasks of web applications need to model categorical\nvariables, such as user IDs and demographics like genders and occupations. To\napply standard machine learning techniques, these categorical predictors are\nalways converted to a set of binary features via one-hot encoding, making the\nresultant feature vector highly sparse. To learn from such sparse data\neffectively, it is crucial to account for the interactions between features.\n  Factorization Machines (FMs) are a popular solution for efficiently using the\nsecond-order feature interactions. However, FM models feature interactions in a\nlinear way, which can be insufficient for capturing the non-linear and complex\ninherent structure of real-world data. While deep neural networks have recently\nbeen applied to learn non-linear feature interactions in industry, such as the\nWide&Deep by Google and DeepCross by Microsoft, the deep structure meanwhile\nmakes them difficult to train.\n  In this paper, we propose a novel model Neural Factorization Machine (NFM)\nfor prediction under sparse settings. NFM seamlessly combines the linearity of\nFM in modelling second-order feature interactions and the non-linearity of\nneural network in modelling higher-order feature interactions. Conceptually,\nNFM is more expressive than FM since FM can be seen as a special case of NFM\nwithout hidden layers. Empirical results on two regression tasks show that with\none hidden layer only, NFM significantly outperforms FM with a 7.3% relative\nimprovement. Compared to the recent deep learning methods Wide&Deep and\nDeepCross, our NFM uses a shallower structure but offers better performance,\nbeing much easier to train and tune in practice.","url_abs":"http://arxiv.org/abs/1708.05027v1","url_pdf":"http://arxiv.org/pdf/1708.05027v1.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":"neural-factorization-machines-for-sparse","repo_url":"https://github.com/hexiangnan/neural_factorization_machine","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"neural-factorization-machines-for-sparse","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":"neural-factorization-machines-for-sparse","repo_url":"https://github.com/jamesdvance/predicting_clicks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"neural-factorization-machines-for-sparse","repo_url":"https://github.com/shenweichen/DeepCTR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"neural-factorization-machines-for-sparse","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":"neural-factorization-machines-for-sparse","repo_url":"https://github.com/xue-pai/FuxiCTR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"link-prediction","task_name":"Link Prediction"}],"methods":[{"method_slug":"wide-deep","method_name":"Wide&Deep"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/link-prediction-on-movielens-25m","task":"Link Prediction","dataset":"MovieLens 25M","model":"NFM","rank_in_archive_order":2,"of":7,"metrics":{"Hits@10":"0.8132","nDCG@10":"0.5347"},"uses_additional_data":false},{"leaderboard":"/sota/link-prediction-on-yelp","task":"Link Prediction","dataset":"Yelp","model":"NFM","rank_in_archive_order":4,"of":9,"metrics":{"HR@10":"0.8595","nDCG@10":"0.6062"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1708.05027","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}