{"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-models-of-interactions-across-sets","title":"Deep Models of Interactions Across Sets","arxiv_id":"1803.02879","date":"2018-03-07","proceeding":"ICML 2018 7","authors":["Jason Hartford","Devon R Graham","Kevin Leyton-Brown","Siamak Ravanbakhsh"],"abstract":"We use deep learning to model interactions across two or more sets of\nobjects, such as user-movie ratings, protein-drug bindings, or ternary\nuser-item-tag interactions. The canonical representation of such interactions\nis a matrix (or a higher-dimensional tensor) with an exchangeability property:\nthe encoding's meaning is not changed by permuting rows or columns. We argue\nthat models should hence be Permutation Equivariant (PE): constrained to make\nthe same predictions across such permutations. We present a parameter-sharing\nscheme and prove that it could not be made any more expressive without\nviolating PE. This scheme yields three benefits. First, we demonstrate\nstate-of-the-art performance on multiple matrix completion benchmarks. Second,\nour models require a number of parameters independent of the numbers of\nobjects, and thus scale well to large datasets. Third, models can be queried\nabout new objects that were not available at training time, but for which\ninteractions have since been observed. In experiments, our models achieved\nsurprisingly good generalization performance on this matrix extrapolation task,\nboth within domains (e.g., new users and new movies drawn from the same\ndistribution used for training) and even across domains (e.g., predicting music\nratings after training on movies).","url_abs":"http://arxiv.org/abs/1803.02879v2","url_pdf":"http://arxiv.org/pdf/1803.02879v2.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-models-of-interactions-across-sets","repo_url":"https://github.com/mravanba/deep_exchangeable_tensors","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"collaborative-filtering","task_name":"Collaborative Filtering"},{"task_slug":"matrix-completion","task_name":"Matrix Completion"},{"task_slug":"recommendation-systems","task_name":"Recommendation Systems"},{"task_slug":"tag","task_name":"TAG"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/recommendation-systems-on-douban-monti","task":"Recommendation Systems","dataset":"Douban Monti","model":"Factorized EAE","rank_in_archive_order":6,"of":8,"metrics":{"RMSE":"0.738"},"uses_additional_data":false},{"leaderboard":"/sota/recommendation-systems-on-flixster-monti","task":"Recommendation Systems","dataset":"Flixster Monti","model":"Factorized EAE","rank_in_archive_order":4,"of":7,"metrics":{"RMSE":"0.908"},"uses_additional_data":false},{"leaderboard":"/sota/collaborative-filtering-on-movielens-100k","task":"Recommendation Systems","dataset":"MovieLens 100K","model":"Self-Supervised Exchangeable Model","rank_in_archive_order":10,"of":18,"metrics":{"RMSE (u1 Splits)":"0.91"},"uses_additional_data":false},{"leaderboard":"/sota/collaborative-filtering-on-movielens-100k","task":"Recommendation Systems","dataset":"MovieLens 100K","model":"Factorized EAE","rank_in_archive_order":12,"of":18,"metrics":{"RMSE (u1 Splits)":"0.920"},"uses_additional_data":false},{"leaderboard":"/sota/collaborative-filtering-on-movielens-1m","task":"Recommendation Systems","dataset":"MovieLens 1M","model":"Factorized EAE","rank_in_archive_order":15,"of":31,"metrics":{"RMSE":"0.860"},"uses_additional_data":false},{"leaderboard":"/sota/recommendation-systems-on-yahoomusic-monti","task":"Recommendation Systems","dataset":"YahooMusic Monti","model":"Factorized EAE","rank_in_archive_order":3,"of":6,"metrics":{"RMSE":"20.0"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1803.02879","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}