[ { "key": "GELRAGII", "version": 237, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/GELRAGII", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/GELRAGII", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Zhou et al.", "parsedDate": "2008", "numChildren": 0 }, "data": { "key": "GELRAGII", "version": 237, "itemType": "conferencePaper", "title": "Large-Scale Parallel Collaborative Filtering for the Netflix Prize", "creators": [ { "creatorType": "author", "firstName": "Yunhong", "lastName": "Zhou" }, { "creatorType": "author", "firstName": "Dennis", "lastName": "Wilkinson" }, { "creatorType": "author", "firstName": "Robert", "lastName": "Schreiber" }, { "creatorType": "author", "firstName": "Rong", "lastName": "Pan" } ], "abstractNote": "Many recommendation systems suggest items to users by utilizing the\ntechniques of collaborative filtering (CF) based on historical records of\nitems that the users have viewed, purchased, or rated. Two major problems\nthat most CF approaches have to contend with are scalability and\nsparseness of the user profiles. To tackle these issues, in this paper, we\ndescribe a CF algorithm alternating-least-squares with\nweighted-λ-regularization (ALS-WR), which is implemented on a parallel\nMatlab platform. We show empirically that the performance of ALS-WR (in\nterms of root mean squared error (RMSE)) monotonically improves with both\nthe number of features and the number of ALS iterations. We applied the\nALS-WR algorithm on a large-scale CF problem, the Netflix Challenge, with\n1000 hidden features and obtained a RMSE score of 0.8985, which is one of\nthe best results based on a pure method. In addition, combining with the\nparallel version of other known methods, we achieved a performance\nimprovement of 5.91% over Netflix’s own CineMatch recommendation system.\nOur method is simple and scales well to very large datasets.", "proceedingsTitle": "", "conferenceName": "Algorithmic Aspects in Information and Management", "publisher": "Springer Berlin Heidelberg", "place": "", "date": "2008", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "337-348", "series": "", "seriesNumber": "", "DOI": "10.1007/978-3-540-68880-8_32", "ISBN": "", "citationKey": "NetflixALS", "url": "http://dx.doi.org/10.1007/978-3-540-68880-8_32", "accessDate": "", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [ { "tag": "LensKit References" } ], "collections": [ "9JMHQD9K" ], "relations": { "dc:replaces": "http://zotero.org/groups/271074/items/HPBPVKAW" }, "dateAdded": "2019-05-30T14:28:20Z", "dateModified": "2026-03-13T17:30:17Z" } }, { "key": "Q9Y2UNCM", "version": 236, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/Q9Y2UNCM", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/Q9Y2UNCM", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Funk", "parsedDate": "2006-12-11", "numChildren": 0 }, "data": { "key": "Q9Y2UNCM", "version": 236, "itemType": "webpage", "title": "Netflix Update: Try This at Home", "creators": [ { "creatorType": "author", "firstName": "Simon", "lastName": "Funk" } ], "abstractNote": "", "websiteTitle": "", "websiteType": "", "date": "December 11, 2006", "publisher": "", "place": "", "DOI": "", "citationKey": "FunkSVD", "url": "http://sifter.org/~simon/journal/20061211.html", "accessDate": "2010-04-08", "shortTitle": "", "language": "", "rights": "", "extra": "", "tags": [ { "tag": "Zotero Import (Mar 30)" }, { "tag": "Zotero Import (Mar 30)/My Library" }, { "tag": "Zotero Import (Mar 30)/My Library/Eval Grant" }, { "tag": "Zotero Import (Mar 30)/My Library/LensKit" }, { "tag": "Zotero Import (Mar 30)/My Library/Recommender Systems" }, { "tag": "Zotero Import (Mar 30)/My Library/Recommender Systems/Error Analysis Paper" }, { "tag": "Zotero Import (Mar 30)/My Library/Recommender Systems/List Comparison Paper" }, { "tag": "Zotero Import (Mar 30)/My Library/Thesis" } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2019-05-30T14:28:26Z", "dateModified": "2026-03-13T17:26:30Z" } }, { "key": "2SI8NQ6R", "version": 235, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/2SI8NQ6R", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/2SI8NQ6R", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Pilászy et al.", "parsedDate": "2010", "numChildren": 0 }, "data": { "key": "2SI8NQ6R", "version": 235, "itemType": "conferencePaper", "title": "Fast ALS-based Matrix Factorization for Explicit and Implicit Feedback Datasets", "creators": [ { "creatorType": "author", "firstName": "István", "lastName": "Pilászy" }, { "creatorType": "author", "firstName": "Dávid", "lastName": "Zibriczky" }, { "creatorType": "author", "firstName": "Domonkos", "lastName": "Tikk" } ], "abstractNote": "Alternating least squares (ALS) is a powerful matrix factorization (MF)\nalgorithm for both explicit and implicit feedback based recommender\nsystems. As shown in many articles, increasing the number of latent\nfactors (denoted by K) boosts the prediction accuracy of MF based\nrecommender systems, including ALS as well. The price of the better\naccuracy is paid by the increased running time: the running time of the\noriginal version of ALS is proportional to K3. Yet, the running time of\nmodel building can be important in recommendation systems; if the model\ncannot keep up with the changing item portfolio and/or user profile, the\nprediction accuracy can be degraded. In this paper we present novel and\nfast ALS variants both for the implicit and explicit feedback datasets,\nwhich offers better trade-off between running time and accuracy. Due to\nthe significantly lower computational complexity of the algorithm - linear\nin terms of K - the model being generated under the same amount of time is\nmore accurate, since the faster training enables to build model with more\nlatent factors. We demonstrate the efficiency of our ALS variants on two\ndatasets using two performance measures, RMSE and average relative\nposition (ARP), and show that either a significantly more accurate model\ncan be generated under the same amount of time or a model with similar\nprediction accuracy can be created faster; for explicit feedback the\nspeed-up factor can be even 5-10.", "proceedingsTitle": "RecSys '10", "conferenceName": "Proceedings of the Fourth ACM Conference on Recommender Systems", "publisher": "ACM", "place": "New York, NY, USA", "date": "2010", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "71–78", "series": "", "seriesNumber": "", "DOI": "10.1145/1864708.1864726", "ISBN": "", "citationKey": "ExplicitALS", "url": "http://doi.acm.org/10.1145/1864708.1864726", "accessDate": "2015-06-01", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "Journal Abbreviation: RecSys '10", "tags": [ { "tag": "CAREER" }, { "tag": "LensKit References" }, { "tag": "alternating least squares", "type": 1 }, { "tag": "collaborative filtering", "type": 1 }, { "tag": "computational complexity", "type": 1 }, { "tag": "implicit and explicit feedback", "type": 1 }, { "tag": "matrix factorization", "type": 1 }, { "tag": "ridge regression", "type": 1 } ], "collections": [ "9JMHQD9K" ], "relations": { "dc:replaces": "http://zotero.org/groups/271074/items/2ZRSNPJM" }, "dateAdded": "2019-07-12T22:38:52Z", "dateModified": "2026-03-13T17:26:04Z" } }, { "key": "3998ZTS8", "version": 234, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/3998ZTS8", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/3998ZTS8", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Koren et al.", "parsedDate": "2009-08", "numChildren": 0 }, "data": { "key": "3998ZTS8", "version": 234, "itemType": "journalArticle", "title": "Matrix Factorization Techniques for Recommender Systems", "creators": [ { "creatorType": "author", "firstName": "Y", "lastName": "Koren" }, { "creatorType": "author", "firstName": "R", "lastName": "Bell" }, { "creatorType": "author", "firstName": "C", "lastName": "Volinsky" } ], "abstractNote": "As the Netflix Prize competition has demonstrated, matrix factorization\nmodels are superior to classic nearest neighbor techniques for producing\nproduct recommendations, allowing the incorporation of additional\ninformation such as implicit feedback, temporal effects, and confidence\nlevels.", "publicationTitle": "Computer", "publisher": "", "place": "", "date": "2009-08", "volume": "42", "issue": "8", "section": "", "partNumber": "", "partTitle": "", "pages": "30-37", "series": "", "seriesTitle": "", "seriesText": "", "journalAbbreviation": "", "DOI": "10.1109/MC.2009.263", "citationKey": "KorenMF", "url": "http://dx.doi.org/10.1109/MC.2009.263", "accessDate": "", "PMID": "", "PMCID": "", "ISSN": "0018-9162", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [ { "tag": "CAREER" }, { "tag": "Fair Info Access Paper" }, { "tag": "LensKit References" } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2023-05-19T17:22:02Z", "dateModified": "2026-03-13T17:24:41Z" } }, { "key": "NM2NWIF3", "version": 233, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/NM2NWIF3", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/NM2NWIF3", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "He et al.", "parsedDate": "2020-07-25", "numChildren": 0 }, "data": { "key": "NM2NWIF3", "version": 233, "itemType": "conferencePaper", "title": "LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation", "creators": [ { "creatorType": "author", "firstName": "Xiangnan", "lastName": "He" }, { "creatorType": "author", "firstName": "Kuan", "lastName": "Deng" }, { "creatorType": "author", "firstName": "Xiang", "lastName": "Wang" }, { "creatorType": "author", "firstName": "Yan", "lastName": "Li" }, { "creatorType": "author", "firstName": "YongDong", "lastName": "Zhang" }, { "creatorType": "author", "firstName": "Meng", "lastName": "Wang" } ], "abstractNote": "Graph Convolution Network (GCN) has become new state-ofthe-art for collaborative filtering. Nevertheless, the reasons of its effectiveness for recommendation are not well understood. Existing work that adapts GCN to recommendation lacks thorough ablation analyses on GCN, which is originally designed for graph classification tasks and equipped with many neural network operations. However, we empirically find that the two most common designs in GCNs — feature transformation and nonlinear activation — contribute little to the performance of collaborative filtering. Even worse, including them adds to the difficulty of training and degrades recommendation performance.", "proceedingsTitle": "Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval", "conferenceName": "SIGIR '20: The 43rd International ACM SIGIR conference on research and development in Information Retrieval", "publisher": "ACM", "place": "Virtual Event China", "date": "2020-07-25", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "639-648", "series": "", "seriesNumber": "", "DOI": "10.1145/3397271.3401063", "ISBN": "978-1-4503-8016-4", "citationKey": "LightGCN", "url": "https://dl.acm.org/doi/10.1145/3397271.3401063", "accessDate": "2025-02-26T15:06:52Z", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "LightGCN", "language": "en", "libraryCatalog": "DOI.org (Crossref)", "callNumber": "", "rights": "", "extra": "", "tags": [ { "tag": "to-read" } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-08-26T18:04:41Z", "dateModified": "2026-03-13T16:00:21Z" } }, { "key": "J2TUASCD", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/J2TUASCD", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/J2TUASCD", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Steck", "parsedDate": "2019-05-13", "numChildren": 0 }, "data": { "key": "J2TUASCD", "version": 232, "itemType": "conferencePaper", "title": "Embarrassingly Shallow Autoencoders for Sparse Data", "creators": [ { "creatorType": "author", "firstName": "Harald", "lastName": "Steck" } ], "abstractNote": "Combining simple elements from the literature, we define a linear model that is geared toward sparse data, in particular implicit feedback data for recommender systems. We show that its training objective has a closed-form solution, and discuss the resulting conceptual insights. Surprisingly, this simple model achieves better ranking accuracy than various state-of-the-art collaborative-filtering approaches, including deep non-linear models, on most of the publicly available data-sets used in our experiments.", "proceedingsTitle": "The World Wide Web Conference", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "May 13, 2019", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "3251–3257", "series": "WWW '19", "seriesNumber": "", "DOI": "10.1145/3308558.3313710", "ISBN": "978-1-4503-6674-8", "citationKey": "steckEmbarrassinglyShallowAutoencoders2019", "url": "https://dl.acm.org/doi/10.1145/3308558.3313710", "accessDate": "2026-02-11", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2026-02-11T22:49:15Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "FLCGQDKU", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/FLCGQDKU", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/FLCGQDKU", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Breese et al.", "parsedDate": "1998", "numChildren": 0 }, "data": { "key": "FLCGQDKU", "version": 232, "itemType": "conferencePaper", "title": "Empirical Analysis of Predictive Algorithms for Collaborative Filtering", "creators": [ { "creatorType": "author", "firstName": "John S", "lastName": "Breese" }, { "creatorType": "author", "firstName": "David", "lastName": "Heckerman" }, { "creatorType": "author", "firstName": "Carl", "lastName": "Kadie" } ], "abstractNote": "Collaborative filtering or recommender systems use a database about user\npreferences to predict additional topics or products a new user might\nlike. In this paper we describe several algorithms designed for this task,\nincluding techniques based on correlation coefficients, vector-based\nsimilarity calculations, and statistical Bayesian methods. We compare the\npredictive accuracy of the various methods in a set of representative\nproblem domains. We use two basic classes of evaluation metrics. The first\ncharacterizes accuracy over a set of individual predictions in terms of\naverage absolute deviation. The second estimates the utility of a ranked\nlist of suggested items. This metric uses an estimate of the probability\nthat a user will see a recommendation in an ordered list. Experiments were\nrun for datasets associated with 3 application areas, 4 experimental\nprotocols, and the 2 evaluation metr rics for the various algorithms.\nResults indicate that for a wide range of conditions, Bayesian networks\nwith decision trees at each node and correlation methods outperform\nBayesian-clustering and vector-similarity methods. Between correlation and\nBayesian networks, the preferred method depends on the nature of the\ndataset, nature of the application (ranked versus one-by-one\npresentation), and the availability of votes with which to make\npredictions. Other considerations include the size of database, speed of\npredictions, and learning time.", "proceedingsTitle": "", "conferenceName": "Proceedings of the Fourteenth Conference on Uncertainty in Artificial Intelligence", "publisher": "", "place": "", "date": "1998", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "43–52", "series": "UAI '98", "seriesNumber": "", "DOI": "", "ISBN": "", "citationKey": "breeseEmpiricalAnalysisPredictive1998", "url": "http://dl.acm.org/citation.cfm?id=2074100", "accessDate": "2010-08-16", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [ { "tag": "⛔ No DOI found", "type": 1 } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-11-11T21:52:25Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "DX2ASGNM", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/DX2ASGNM", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/DX2ASGNM", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Assylbekov et al.", "parsedDate": "2023-09-14", "numChildren": 0 }, "data": { "key": "DX2ASGNM", "version": 232, "itemType": "conferencePaper", "title": "Delivery Hero Recommendation Dataset: A Novel Dataset for Benchmarking Recommendation Algorithms", "creators": [ { "creatorType": "author", "firstName": "Yernat", "lastName": "Assylbekov" }, { "creatorType": "author", "firstName": "Raghav", "lastName": "Bali" }, { "creatorType": "author", "firstName": "Luke", "lastName": "Bovard" }, { "creatorType": "author", "firstName": "Christian", "lastName": "Klaue" } ], "abstractNote": "In this paper we propose Delivery Hero Recommendation Dataset (DHRD), a novel real-world dataset for researchers. DHRD comprises over a million food delivery orders from three distinct cities, encompassing thousands of vendors and an extensive range of dishes, serving a combined customer base of over a million individuals. We discuss the challenges associated with such real-world datasets. By releasing DHRD, researchers are empowered with a valuable resource for building and evaluating recommender systems, paving the way for advancements in this domain.", "proceedingsTitle": "Proceedings of the 17th ACM Conference on Recommender Systems", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "September 14, 2023", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "1042–1044", "series": "RecSys '23", "seriesNumber": "", "DOI": "10.1145/3604915.3610242", "ISBN": "979-8-4007-0241-9", "citationKey": "assylbekovDeliveryHeroRecommendation2023", "url": "https://dl.acm.org/doi/10.1145/3604915.3610242", "accessDate": "2025-11-04", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "Delivery Hero Recommendation Dataset", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-11-04T15:45:11Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "H9TXUJXM", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/H9TXUJXM", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/H9TXUJXM", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Schubert and Gertz", "parsedDate": "2018-07-09", "numChildren": 0 }, "data": { "key": "H9TXUJXM", "version": 232, "itemType": "conferencePaper", "title": "Numerically stable parallel computation of (co-)variance", "creators": [ { "creatorType": "author", "firstName": "Erich", "lastName": "Schubert" }, { "creatorType": "author", "firstName": "Michael", "lastName": "Gertz" } ], "abstractNote": "With the advent of big data, we see an increasing interest in computing correlations in huge data sets with both many instances and many variables. Essential descriptive statistics such as the variance, standard deviation, covariance, and correlation can suffer from a numerical instability known as \"catastrophic cancellation\" that can lead to problems when naively computing these statistics with a popular textbook equation. While this instability has been discussed in the literature already 50 years ago, we found that even today, some high-profile tools still employ the instable version.In this paper, we study a popular incremental technique originally proposed by Welford, which we extend to weighted covariance and correlation. We also discuss strategies for further improving numerical precision, how to compute such statistics online on a data stream, with exponential aging, with missing data, and a batch parallelization for both high performance and numerical precision.We demonstrate when the numerical instability arises, and the performance of different approaches under these conditions. We showcase applications from the classic computation of variance as well as advanced applications such as stock market analysis with exponentially weighted moving models and Gaussian mixture modeling for cluster analysis that all benefit from this approach.", "proceedingsTitle": "Proceedings of the 30th International Conference on Scientific and Statistical Database Management", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "July 9, 2018", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "1–12", "series": "SSDBM '18", "seriesNumber": "", "DOI": "10.1145/3221269.3223036", "ISBN": "978-1-4503-6505-5", "citationKey": "schubertNumericallyStableParallel2018", "url": "https://dl.acm.org/doi/10.1145/3221269.3223036", "accessDate": "2025-10-23", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-10-23T23:51:52Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "IUWPCQN3", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/IUWPCQN3", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/IUWPCQN3", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Rendle and Freudenthaler", "parsedDate": "2014-02-24", "numChildren": 0 }, "data": { "key": "IUWPCQN3", "version": 232, "itemType": "conferencePaper", "title": "Improving pairwise learning for item recommendation from implicit feedback", "creators": [ { "creatorType": "author", "firstName": "Steffen", "lastName": "Rendle" }, { "creatorType": "author", "firstName": "Christoph", "lastName": "Freudenthaler" } ], "abstractNote": "Pairwise algorithms are popular for learning recommender systems from\nimplicit feedback. For each user, or more generally context, they try to\ndiscriminate between a small set of selected items and the large set of\nremaining (irrelevant) items. Learning is typically based on stochastic\ngradient descent (SGD) with uniformly drawn pairs. In this work, we show\nthat convergence of such SGD learning algorithms slows down considerably\nif the item popularity has a tailed distribution. We propose a non-uniform\nitem sampler to overcome this problem. The proposed sampler is\ncontext-dependent and oversamples informative pairs to speed up\nconvergence. An efficient implementation with constant amortized runtime\ncosts is developed. Furthermore, it is shown how the proposed learning\nalgorithm can be applied to a large class of recommender models. The\nproperties of the new learning algorithm are studied empirically on two\nreal-world recommender system problems. The experiments indicate that the\nproposed adaptive sampler improves the state-of-the art learning algorithm\nlargely in convergence without negative effects on prediction quality or\niteration runtime.", "proceedingsTitle": "WSDM '14", "conferenceName": "Proceedings of the 7th ACM international conference on Web search and data mining", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "2014-02-24", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "273-282", "series": "", "seriesNumber": "", "DOI": "10.1145/2556195.2556248", "ISBN": "", "citationKey": "rendleImprovingPairwiseLearning2014", "url": "https://doi.org/10.1145/2556195.2556248", "accessDate": "2021-02-23", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "Journal Abbreviation: WSDM '14", "tags": [ { "tag": "factorization model", "type": 1 }, { "tag": "item recommendation", "type": 1 }, { "tag": "matrix factorization", "type": 1 }, { "tag": "recommender systems", "type": 1 } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-10-21T19:14:47Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "IDM6EX3H", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/IDM6EX3H", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/IDM6EX3H", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Ning and Karypis", "parsedDate": "2011", "numChildren": 0 }, "data": { "key": "IDM6EX3H", "version": 232, "itemType": "conferencePaper", "title": "SLIM: Sparse Linear Methods for Top-N Recommender Systems", "creators": [ { "creatorType": "author", "firstName": "Xia", "lastName": "Ning" }, { "creatorType": "author", "firstName": "George", "lastName": "Karypis" } ], "abstractNote": "This paper focuses on developing effective and efficient algorithms for\ntop-N recommender systems. A novel Sparse Linear Method (SLIM) is\nproposed, which generates top-N recommendations by aggregating from user\npurchase/rating profiles. A sparse aggregation coefficient matrix W is\nlearned from SLIM by solving an `1-norm and `2-norm regularized\noptimization problem. W is demonstrated to produce high quality\nrecommendations and its sparsity allows SLIM to generate recommendations\nvery fast. A comprehensive set of experiments is conducted by comparing\nthe SLIM method and other state-of-the-art top-N recommendation methods.\nThe experiments show that SLIM achieves significant improvements both in\nrun time performance and recommendation quality over the best existing\nmethods.", "proceedingsTitle": "ICDM '11", "conferenceName": "Proceedings of the 2011 IEEE 11th International Conference on Data Mining", "publisher": "IEEE Computer Society", "place": "Washington, DC, USA", "date": "2011", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "497–506", "series": "", "seriesNumber": "", "DOI": "10.1109/ICDM.2011.134", "ISBN": "", "citationKey": "ningSLIMSparseLinear2011a", "url": "http://dx.doi.org/10.1109/ICDM.2011.134", "accessDate": "2017-01-04", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "Journal Abbreviation: ICDM '11", "tags": [], "collections": [], "relations": {}, "dateAdded": "2025-10-14T16:45:29Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "7J2SJF3A", "version": 232, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/7J2SJF3A", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/7J2SJF3A", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Webber et al.", "parsedDate": "2010-11-23", "numChildren": 0 }, "data": { "key": "7J2SJF3A", "version": 232, "itemType": "journalArticle", "title": "A similarity measure for indefinite rankings", "creators": [ { "creatorType": "author", "firstName": "William", "lastName": "Webber" }, { "creatorType": "author", "firstName": "Alistair", "lastName": "Moffat" }, { "creatorType": "author", "firstName": "Justin", "lastName": "Zobel" } ], "abstractNote": "Ranked lists are encountered in research and daily life and it is often of interest to compare these lists even when they are incomplete or have only some members in common. An example is document rankings returned for the same query by different search engines. A measure of the similarity between incomplete rankings should handle nonconjointness, weight high ranks more heavily than low, and be monotonic with increasing depth of evaluation; but no measure satisfying all these criteria currently exists. In this article, we propose a new measure having these qualities, namely rank-biased overlap (RBO). The RBO measure is based on a simple probabilistic user model. It provides monotonicity by calculating, at a given depth of evaluation, a base score that is non-decreasing with additional evaluation, and a maximum score that is nonincreasing. An extrapolated score can be calculated between these bounds if a point estimate is required. RBO has a parameter which determines the strength of the weighting to top ranks. We extend RBO to handle tied ranks and rankings of different lengths. Finally, we give examples of the use of the measure in comparing the results produced by public search engines and in assessing retrieval systems in the laboratory.", "publicationTitle": "ACM Trans. Inf. Syst.", "publisher": "", "place": "", "date": "November 23, 2010", "volume": "28", "issue": "4", "section": "", "partNumber": "", "partTitle": "", "pages": "20:1–20:38", "series": "", "seriesTitle": "", "seriesText": "", "journalAbbreviation": "", "DOI": "10.1145/1852102.1852106", "citationKey": "webberSimilarityMeasureIndefinite2010", "url": "https://doi.org/10.1145/1852102.1852106", "accessDate": "2025-10-14T15:35:40Z", "PMID": "", "PMCID": "", "ISSN": "1046-8188", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-10-14T15:35:40Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "72KVVIAT", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/72KVVIAT", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/72KVVIAT", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Isufaj et al.", "parsedDate": "2025-09-07", "numChildren": 0 }, "data": { "key": "72KVVIAT", "version": 231, "itemType": "conferencePaper", "title": "The XITE Million Sessions Dataset", "creators": [ { "creatorType": "author", "firstName": "Ralvi", "lastName": "Isufaj" }, { "creatorType": "author", "firstName": "Ruslan", "lastName": "Tsygankov" }, { "creatorType": "author", "firstName": "Zoltán", "lastName": "Szlávik" } ], "abstractNote": "We present the XITE Million Sessions Dataset, a collection of one million music video streaming sessions from an interactive TV platform. This dataset addresses a significant gap in music recommendation research by capturing sequential user interactions with music video content. Each session contains sequences of videos watched by anonymised users, along with metadata including artist information, title, genre and subgenre classifications from XITE’s expert-curated taxonomy, and watch-time metrics. The dataset also includes XITE’s genre hierarchy and subgenre correlation matrix, representing musical relationships established by music experts. We provide MusicBrainz identifiers where possible to enable connections with external music resources. While we do not include the video content itself, the dataset documents how users engage with music in a video-based environment, which may exhibit interaction patterns that differ from audio-only consumption. To demonstrate the dataset’s research utility, we benchmark a standard playlist continuation task using transformer-based and graph-based models. This contribution allows researchers to develop and evaluate recommendation algorithms for music video consumption and examine how existing methods generalise beyond audio-only datasets to screen-based music experiences.", "proceedingsTitle": "Proceedings of the Nineteenth ACM Conference on Recommender Systems", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "September 7, 2025", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "857–864", "series": "RecSys '25", "seriesNumber": "", "DOI": "10.1145/3705328.3748168", "ISBN": "979-8-4007-1364-4", "citationKey": "isufajXITEMillionSessions2025", "url": "https://dl.acm.org/doi/10.1145/3705328.3748168", "accessDate": "2025-09-25", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [ { "tag": "dataset" } ], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-09-25T10:10:52Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "PFAZMWPB", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/PFAZMWPB", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/PFAZMWPB", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Kruse et al.", "parsedDate": "2024-10-14", "numChildren": 0 }, "data": { "key": "PFAZMWPB", "version": 231, "itemType": "conferencePaper", "title": "EB-NeRD a large-scale dataset for news recommendation", "creators": [ { "creatorType": "author", "firstName": "Johannes", "lastName": "Kruse" }, { "creatorType": "author", "firstName": "Kasper", "lastName": "Lindskow" }, { "creatorType": "author", "firstName": "Saikishore", "lastName": "Kalloori" }, { "creatorType": "author", "firstName": "Marco", "lastName": "Polignano" }, { "creatorType": "author", "firstName": "Claudio", "lastName": "Pomo" }, { "creatorType": "author", "firstName": "Abhishek", "lastName": "Srivastava" }, { "creatorType": "author", "firstName": "Anshuk", "lastName": "Uppal" }, { "creatorType": "author", "firstName": "Michael Riis", "lastName": "Andersen" }, { "creatorType": "author", "firstName": "Jes", "lastName": "Frellsen" } ], "abstractNote": "Personalized content recommendations have been pivotal to the content experience in digital media from video streaming to social networks. However, several domain specific challenges have held back adoption of recommender systems in news publishing. To address these challenges, we introduce the Ekstra Bladet News Recommendation Dataset (EB-NeRD). The dataset encompasses data from over a million unique users and more than 37 million impression logs from Ekstra Bladet. It also includes a collection of over 125, 000 Danish news articles, complete with titles, abstracts, bodies, and metadata, such as categories. EB-NeRD served as the benchmark dataset for the RecSys ’24 Challenge, where it was demonstrated how the dataset can be used to address both technical and normative challenges in designing effective and responsible recommender systems for news publishing. The dataset is available at: https://recsys.eb.dk.", "proceedingsTitle": "Proceedings of the Recommender Systems Challenge 2024", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "October 14, 2024", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "1–11", "series": "RecSysChallenge '24", "seriesNumber": "", "DOI": "10.1145/3687151.3687152", "ISBN": "979-8-4007-1127-5", "citationKey": "kruseEBNeRDLargescaleDataset2024", "url": "https://dl.acm.org/doi/10.1145/3687151.3687152", "accessDate": "2025-09-02", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-09-02T15:29:32Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "W9GL5WA8", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/W9GL5WA8", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/W9GL5WA8", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Gao et al.", "parsedDate": "2022-10-17", "numChildren": 0 }, "data": { "key": "W9GL5WA8", "version": 231, "itemType": "conferencePaper", "title": "KuaiRec: A Fully-observed Dataset and Insights for Evaluating Recommender Systems", "creators": [ { "creatorType": "author", "firstName": "Chongming", "lastName": "Gao" }, { "creatorType": "author", "firstName": "Shijun", "lastName": "Li" }, { "creatorType": "author", "firstName": "Wenqiang", "lastName": "Lei" }, { "creatorType": "author", "firstName": "Jiawei", "lastName": "Chen" }, { "creatorType": "author", "firstName": "Biao", "lastName": "Li" }, { "creatorType": "author", "firstName": "Peng", "lastName": "Jiang" }, { "creatorType": "author", "firstName": "Xiangnan", "lastName": "He" }, { "creatorType": "author", "firstName": "Jiaxin", "lastName": "Mao" }, { "creatorType": "author", "firstName": "Tat-Seng", "lastName": "Chua" } ], "abstractNote": "The progress of recommender systems is hampered mainly by evaluation as it requires real-time interactions between humans and systems, which is too laborious and expensive. This issue is usually approached by utilizing the interaction history to conduct offline evaluation. However, existing datasets of user-item interactions are partially observed, leaving it unclear how and to what extent the missing interactions will influence the evaluation. To answer this question, we collect a fully-observed dataset from Kuaishou's online environment, where almost all 1,411 users have been exposed to all 3,327 items. To the best of our knowledge, this is the first real-world fully-observed data with millions of user-item interactions.With this unique dataset, we conduct a preliminary analysis of how the two factors - data density and exposure bias - affect the evaluation results of multi-round conversational recommendation. Our main discoveries are that the performance ranking of different methods varies with the two factors, and this effect can only be alleviated in certain cases by estimating missing interactions for user simulation. This demonstrates the necessity of the fully-observed dataset. We release the dataset and the pipeline implementation for evaluation at https://kuairec.com", "proceedingsTitle": "Proceedings of the 31st ACM International Conference on Information & Knowledge Management", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "October 17, 2022", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "540–550", "series": "CIKM '22", "seriesNumber": "", "DOI": "10.1145/3511808.3557220", "ISBN": "978-1-4503-9236-5", "citationKey": "gaoKuaiRecFullyobservedDataset2022", "url": "https://dl.acm.org/doi/10.1145/3511808.3557220", "accessDate": "2025-08-28", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "KuaiRec", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-08-28T14:02:51Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "3SGZ5MTT", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/3SGZ5MTT", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/3SGZ5MTT", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Kleiner et al.", "parsedDate": "2014-09-01", "numChildren": 0 }, "data": { "key": "3SGZ5MTT", "version": 231, "itemType": "journalArticle", "title": "A scalable bootstrap for massive data", "creators": [ { "creatorType": "author", "firstName": "Ariel", "lastName": "Kleiner" }, { "creatorType": "author", "firstName": "Ameet", "lastName": "Talwalkar" }, { "creatorType": "author", "firstName": "Purnamrita", "lastName": "Sarkar" }, { "creatorType": "author", "firstName": "Michael I.", "lastName": "Jordan" } ], "abstractNote": "The bootstrap provides a simple and powerful means of assessing the quality of estimators. However, in settings involving large data sets—which are increasingly prevalent—the calculation of bootstrap-based quantities can be prohibitively demanding computationally. Although variants such as subsampling and the m out of n bootstrap can be used in principle to reduce the cost of bootstrap computations, these methods are generally not robust to specification of tuning parameters (such as the number of subsampled data points), and they often require knowledge of the estimator's convergence rate, in contrast with the bootstrap. As an alternative, we introduce the ‘bag of little bootstraps’ (BLB), which is a new procedure which incorporates features of both the bootstrap and subsampling to yield a robust, computationally efficient means of assessing the quality of estimators. The BLB is well suited to modern parallel and distributed computing architectures and furthermore retains the generic applicability and statistical efficiency of the bootstrap. We demonstrate the BLB's favourable statistical performance via a theoretical analysis elucidating the procedure's properties, as well as a simulation study comparing the BLB with the bootstrap, the m out of n bootstrap and subsampling. In addition, we present results from a large-scale distributed implementation of the BLB demonstrating its computational superiority on massive data, a method for adaptively selecting the BLB's tuning parameters, an empirical study applying the BLB to several real data sets and an extension of the BLB to time series data.", "publicationTitle": "Journal of the Royal Statistical Society Series B: Statistical Methodology", "publisher": "", "place": "", "date": "2014-09-01", "volume": "76", "issue": "4", "section": "", "partNumber": "", "partTitle": "", "pages": "795-816", "series": "", "seriesTitle": "", "seriesText": "", "journalAbbreviation": "Journal of the Royal Statistical Society Series B: Statistical Methodology", "DOI": "10.1111/rssb.12050", "citationKey": "kleinerScalableBootstrapMassive2014", "url": "https://doi.org/10.1111/rssb.12050", "accessDate": "2025-05-30T21:38:12Z", "PMID": "", "PMCID": "", "ISSN": "1369-7412", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "Silverchair", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-05-30T21:38:19Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "YIFN34PI", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/YIFN34PI", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/YIFN34PI", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Rendle et al.", "parsedDate": "2020-09-22", "numChildren": 0 }, "data": { "key": "YIFN34PI", "version": 231, "itemType": "conferencePaper", "title": "Neural collaborative filtering vs. matrix factorization revisited", "creators": [ { "creatorType": "author", "firstName": "Steffen", "lastName": "Rendle" }, { "creatorType": "author", "firstName": "Walid", "lastName": "Krichene" }, { "creatorType": "author", "firstName": "Li", "lastName": "Zhang" }, { "creatorType": "author", "firstName": "John", "lastName": "Anderson" } ], "abstractNote": "Embedding based models have been the state of the art in collaborative\nfiltering for over a decade. Traditionally, the dot product or higher\norder equivalents have been used to combine two or more embeddings, e.g.,\nmost notably in matrix factorization. In recent years, it was suggested to\nreplace the dot product with a learned similarity e.g. using a multilayer\nperceptron (MLP). This approach is often referred to as neural\ncollaborative filtering (NCF). In this work, we revisit the experiments of\nthe NCF paper that popularized learned similarities using MLPs. First, we\nshow that with a proper hyperparameter selection, a simple dot product\nsubstantially outperforms the proposed learned similarities. Second, while\na MLP can in theory approximate any function, we show that it is\nnon-trivial to learn a dot product with an MLP. Finally, we discuss\npractical issues that arise when applying MLP based similarities and show\nthat MLPs are too costly to use for item recommendation in production\nenvironments while dot products allow to apply very efficient retrieval\nalgorithms. We conclude that MLPs should be used with care as embedding\ncombiner and that dot products might be a better default choice.", "proceedingsTitle": "RecSys '20", "conferenceName": "Fourteenth ACM Conference on Recommender Systems", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "2020-09-22", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "240-248", "series": "", "seriesNumber": "", "DOI": "10.1145/3383313.3412488", "ISBN": "", "citationKey": "rendleNeuralCollaborativeFiltering2020", "url": "https://doi.org/10.1145/3383313.3412488", "accessDate": "2020-09-20", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "Journal Abbreviation: RecSys '20", "tags": [ { "tag": "Item Recommendation", "type": 1 }, { "tag": "Matrix Factorization", "type": 1 }, { "tag": "Neural Collaborative Filtering", "type": 1 } ], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-04-25T15:32:58Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "P8HUNA2S", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/P8HUNA2S", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/P8HUNA2S", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Jannach et al.", "parsedDate": "2022", "numChildren": 0 }, "data": { "key": "P8HUNA2S", "version": 231, "itemType": "bookSection", "title": "Session-based recommender systems", "creators": [ { "creatorType": "author", "firstName": "Dietmar", "lastName": "Jannach" }, { "creatorType": "author", "firstName": "Massimo", "lastName": "Quadrana" }, { "creatorType": "author", "firstName": "Paolo", "lastName": "Cremonesi" }, { "creatorType": "editor", "firstName": "Francesco", "lastName": "Ricci" }, { "creatorType": "editor", "firstName": "Lior", "lastName": "Rokach" }, { "creatorType": "editor", "firstName": "Bracha", "lastName": "Shapira" } ], "abstractNote": "Session-based recommendation is concerned with the problem of tailoring\nitem suggestions according to the short-term needs and assumed intents of\nthe user.", "bookTitle": "Recommender Systems Handbook", "series": "", "seriesNumber": "", "volume": "", "numberOfVolumes": "", "edition": "", "date": "2022", "publisher": "Springer US", "place": "New York, NY", "originalDate": "", "originalPublisher": "", "originalPlace": "", "format": "", "pages": "301-334", "ISBN": "978-1-0716-2197-4", "DOI": "10.1007/978-1-0716-2197-4_8", "citationKey": "jannachSessionbasedRecommenderSystems2022", "url": "https://doi.org/10.1007/978-1-0716-2197-4_8", "accessDate": "", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-04-25T15:31:53Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "GWJ4V5N7", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/GWJ4V5N7", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/GWJ4V5N7", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Wegmeth et al.", "parsedDate": "2024-09-23", "numChildren": 0 }, "data": { "key": "GWJ4V5N7", "version": 231, "itemType": "preprint", "title": "EMERS: energy meter for recommender systems", "creators": [ { "creatorType": "author", "firstName": "Lukas", "lastName": "Wegmeth" }, { "creatorType": "author", "firstName": "Tobias", "lastName": "Vente" }, { "creatorType": "author", "firstName": "Alan", "lastName": "Said" }, { "creatorType": "author", "firstName": "Joeran", "lastName": "Beel" } ], "abstractNote": "Due to recent advancements in machine learning, recommender systems use increasingly more energy for training, evaluation, and deployment. However, the recommender systems community often does not report the energy consumption of their experiments. In today's research landscape, no tools exist to easily measure the energy consumption of recommender systems experiments. To bridge this gap, we introduce EMERS, the first software library that simplifies measuring, monitoring, recording, and sharing the energy consumption of recommender systems experiments. EMERS measures energy consumption with smart power plugs and offers a user interface to monitor and compare the energy consumption of recommender systems experiments. Thereby, EMERS improves sustainability awareness and simplifies self-reporting energy consumption for recommender systems practitioners and researchers.", "genre": "", "repository": "", "archiveID": "arXiv:2409.15060", "place": "", "date": "2024-09-23", "series": "", "seriesNumber": "", "DOI": "10.48550/arXiv.2409.15060", "citationKey": "wegmethEMERSEnergyMeter2024", "url": "http://arxiv.org/abs/2409.15060", "accessDate": "2024-09-26T12:01:20Z", "archive": "", "archiveLocation": "", "shortTitle": "EMERS", "language": "", "libraryCatalog": "arXiv.org", "callNumber": "", "rights": "", "extra": "arXiv:2409.15060 [cs]", "tags": [ { "tag": "Computer Science - Information Retrieval", "type": 1 } ], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-04-25T15:31:33Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "9NIYT9L4", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/9NIYT9L4", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/9NIYT9L4", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Vente et al.", "parsedDate": "2024-10-14", "numChildren": 1 }, "data": { "key": "9NIYT9L4", "version": 231, "itemType": "conferencePaper", "title": "From clicks to carbon: the environmental toll of recommender systems", "creators": [ { "creatorType": "author", "firstName": "Tobias", "lastName": "Vente" }, { "creatorType": "author", "firstName": "Lukas", "lastName": "Wegmeth" }, { "creatorType": "author", "firstName": "Alan", "lastName": "Said" }, { "creatorType": "author", "firstName": "Joeran", "lastName": "Beel" } ], "abstractNote": "As global warming soars, evaluating the environmental impact of research is more critical now than ever before. However, we find that few to no recommender systems research papers document their impact on the environment. Consequently, in this paper, we conduct a comprehensive analysis of the environmental impact of recommender system research by reproducing a characteristic recommender systems experimental pipeline. We focus on estimating the carbon footprint of recommender systems research papers, highlighting the evolution of the environmental impact of recommender systems research experiments over time. We thoroughly evaluated all 79 full papers from the ACM RecSys conference in the years 2013 and 2023 to analyze representative experimental pipelines for papers utilizing traditional, so-called good old-fashioned AI algorithms and deep learning algorithms, respectively. We reproduced these representative experimental pipelines, measured electricity consumption using a hardware energy meter, and converted the measured energy consumption into CO2 equivalents to estimate the environmental impact. Our results show that a recommender systems research paper utilizing deep learning algorithms emits approximately 42 times more CO2 equivalents than a paper utilizing traditional algorithms. Furthermore, on average, such a paper produces 3,297 kilograms of CO2 equivalents, which is more than one person produces by flying from New York City to Melbourne or the amount one tree sequesters in 300 years.", "proceedingsTitle": "Proceedings of the 18th ACM Conference on Recommender Systems", "conferenceName": "RecSys '24", "publisher": "ACM", "place": "", "date": "2024-10-14", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "", "series": "", "seriesNumber": "", "DOI": "10.1145/3640457.203688074", "ISBN": "", "citationKey": "venteClicksCarbonEnvironmental2024", "url": "http://arxiv.org/abs/2408.08203", "accessDate": "2024-08-16T12:50:25Z", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "From clicks to carbon", "language": "", "libraryCatalog": "arXiv.org", "callNumber": "", "rights": "", "extra": "arXiv:2408.08203 [cs]", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-04-25T15:31:30Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "VRZU6XEB", "version": 231, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/VRZU6XEB", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/VRZU6XEB", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Vargas and Castells", "parsedDate": "2011", "numChildren": 1 }, "data": { "key": "VRZU6XEB", "version": 231, "itemType": "conferencePaper", "title": "Rank and Relevance in Novelty and Diversity Metrics for Recommender Systems", "creators": [ { "creatorType": "author", "firstName": "Saúl", "lastName": "Vargas" }, { "creatorType": "author", "firstName": "Pablo", "lastName": "Castells" } ], "abstractNote": "The Recommender Systems community is paying increasing attention to\nnovelty and diversity as key qualities beyond accuracy in real\nrecommendation scenarios. Despite the raise of interest and work on the\ntopic in recent years, we find that a clear common methodological and\nconceptual ground for the evaluation of these dimensions is still to be\nconsolidated. Different evaluation metrics have been reported in the\nliterature but the precise relation, distinction or equivalence between\nthem has not been explicitly studied. Furthermore, the metrics reported so\nfar miss important properties such as taking into consideration the\nranking of recommended items, or whether items are relevant or not, when\nassessing the novelty and diversity of recommendations. We present a\nformal framework for the definition of novelty and diversity metrics that\nunifies and generalizes several state of the art metrics. We identify\nthree essential ground concepts at the roots of novelty and diversity:\nchoice, discovery and relevance, upon which the framework is built. Item\nrank and relevance are introduced through a probabilistic recommendation\nbrowsing model, building upon the same three basic concepts. Based on the\ncombination of ground elements, and the assumptions of the browsing model,\ndifferent metrics and variants unfold. We report experimental observations\nwhich validate and illustrate the properties of the proposed metrics.", "proceedingsTitle": "RecSys '11", "conferenceName": "Proceedings of the Fifth ACM Conference on Recommender Systems", "publisher": "ACM", "place": "New York, NY, USA", "date": "2011", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "109–116", "series": "", "seriesNumber": "", "DOI": "10.1145/2043932.2043955", "ISBN": "", "citationKey": "vargasRankRelevanceNovelty2011", "url": "http://doi.acm.org/10.1145/2043932.2043955", "accessDate": "2014-05-03", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "Journal Abbreviation: RecSys '11", "tags": [ { "tag": "toolkit", "type": 1 } ], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-02-04T15:51:29Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "BLZV97J2", "version": 230, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/BLZV97J2", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/BLZV97J2", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Ekstrand and Mahant", "parsedDate": "2017-05-22", "numChildren": 0 }, "data": { "key": "BLZV97J2", "version": 230, "itemType": "conferencePaper", "title": "Sturgeon and the Cool Kids: Problems with Top-N Recommender Evaluation", "creators": [ { "creatorType": "author", "firstName": "Michael D", "lastName": "Ekstrand" }, { "creatorType": "author", "firstName": "Vaibhav", "lastName": "Mahant" } ], "abstractNote": "Top-N evaluation of recommender systems, typically carried out using\nmetrics from information retrieval or machine learning, has several\nchallenges. Two of these challenges are popularity bias, where the\nevaluation intrinsically favors algorithms that recommend popular items,\nand misclassified decoys, where items for which no user relevance is known\nare actually relevant to the user, but the evaluation is unaware and\npenalizes the recommender for suggesting them. One strategy for mitigating\nthe misclassified decoy problem is the one-plus-random evaluation strategy\nand its generalization, which we call random decoys. In this work, we\nexplore the random decoy strategy through both a theoretical treatment and\nan empirical study, but find little evidence to guide its tuning and show\nthat it has complex and deleterious interactions with popularity bias.", "proceedingsTitle": "Proceedings of the 30th Florida Artificial Intelligence Research Society Conference", "conferenceName": "FLAIRS 30", "publisher": "AAAI Press", "place": "", "date": "2017-05-22", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "", "series": "FLAIRS 30", "seriesNumber": "", "DOI": "", "ISBN": "", "citationKey": "ekstrandSturgeonCoolKids2017a", "url": "https://aaai.org/papers/639-flairs-2017-15534/", "accessDate": "", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-02-04T15:50:54Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "5RAR2V3C", "version": 230, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/5RAR2V3C", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/5RAR2V3C", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Vig et al.", "parsedDate": "2012-09", "numChildren": 0 }, "data": { "key": "5RAR2V3C", "version": 230, "itemType": "journalArticle", "title": "The Tag Genome: Encoding Community Knowledge to Support Novel Interaction", "creators": [ { "creatorType": "author", "firstName": "Jesse", "lastName": "Vig" }, { "creatorType": "author", "firstName": "Shilad", "lastName": "Sen" }, { "creatorType": "author", "firstName": "John", "lastName": "Riedl" } ], "abstractNote": "This article introduces the tag genome, a data structure that extends the\ntraditional tagging model to provide enhanced forms of user interaction.\nJust as a biological genome encodes an organism based on a sequence of\ngenes, the tag genome encodes an item in an information space based on its\nrelationship to a common set of tags. We present a machine learning\napproach for computing the tag genome, and we evaluate several learning\nmodels on a ground truth dataset provided by users. We describe an\napplication of the tag genome called Movie Tuner which enables users to\nnavigate from one item to nearby items along dimensions represented by\ntags. We present the results of a 7-week field trial of 2,531 users of\nMovie Tuner and a survey evaluating users’ subjective experience. Finally,\nwe outline the broader space of applications of the tag genome.", "publicationTitle": "ACM Trans. Interact. Intell. Syst.", "publisher": "", "place": "", "date": "2012-09", "volume": "2", "issue": "3", "section": "", "partNumber": "", "partTitle": "", "pages": "13:1–13:44", "series": "", "seriesTitle": "", "seriesText": "", "journalAbbreviation": "", "DOI": "10.1145/2362394.2362395", "citationKey": "vigTagGenomeEncoding2012", "url": "http://doi.acm.org/10.1145/2362394.2362395", "accessDate": "2014-05-02", "PMID": "", "PMCID": "", "ISSN": "2160-6455", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K" ], "relations": {}, "dateAdded": "2025-01-31T22:11:34Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "BCA8IW3Q", "version": 230, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/BCA8IW3Q", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/BCA8IW3Q", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Jeunen et al.", "parsedDate": "2024-08-24", "numChildren": 0 }, "data": { "key": "BCA8IW3Q", "version": 230, "itemType": "conferencePaper", "title": "On (normalised) discounted cumulative gain as an off-policy evaluation metric for top-N recommendation", "creators": [ { "creatorType": "author", "firstName": "Olivier", "lastName": "Jeunen" }, { "creatorType": "author", "firstName": "Ivan", "lastName": "Potapov" }, { "creatorType": "author", "firstName": "Aleksei", "lastName": "Ustimenko" } ], "abstractNote": "Approaches to recommendation are typically evaluated in one of two ways: (1) via a (simulated) online experiment, often seen as the gold standard, or (2) via some offline evaluation procedure, where the goal is to approximate the outcome of an online experiment. Several offline evaluation metrics have been adopted in the literature, inspired by ranking metrics prevalent in the field of Information Retrieval. (Normalised) Discounted Cumulative Gain (nDCG) is one such metric that has seen widespread adoption in empirical studies, and higher (n)DCG values have been used to present new methods as the state-of-the-art in top-n recommendation for many years.Our work takes a critical look at this approach, and investigates when we can expect such metrics to approximate the gold standard outcome of an online experiment. We formally present the assumptions that are necessary to consider DCG an unbiased estimator of online reward and provide a derivation for this metric from first principles, highlighting where we deviate from its traditional uses in IR. Importantly, we show that normalising the metric renders it inconsistent, in that even when DCG is unbiased, ranking competing methods by their normalised DCG can invert their relative order. Through a correlation analysis between off- and on-line experiments conducted on a large-scale recommendation platform, we show that our unbiased DCG estimates strongly correlate with online reward, even when some of the metric's inherent assumptions are violated. This statement no longer holds for its normalised variant, suggesting that nDCG's practical utility may be limited.", "proceedingsTitle": "Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "August 24, 2024", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "1222–1233", "series": "KDD '24", "seriesNumber": "", "DOI": "10.1145/3637528.3671687", "ISBN": "979-8-4007-0490-1", "citationKey": "jeunenNormalisedDiscountedCumulative2024", "url": "https://doi.org/10.1145/3637528.3671687", "accessDate": "2025-01-13", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "ACM Digital Library", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "9JMHQD9K", "LSCSUCXH" ], "relations": {}, "dateAdded": "2025-01-13T15:31:15Z", "dateModified": "2026-02-16T15:58:04Z" } }, { "key": "TNZ42I7U", "version": 230, "library": { "type": "group", "id": 271074, "name": "LensKit", "links": { "alternate": { "href": "https://www.zotero.org/groups/lenskit", "type": "text/html" } } }, "links": { "self": { "href": "https://api.zotero.org/groups/271074/items/TNZ42I7U", "type": "application/json" }, "alternate": { "href": "https://www.zotero.org/groups/lenskit/items/TNZ42I7U", "type": "text/html" } }, "meta": { "createdByUser": { "id": 6655, "username": "ekstrand", "name": "Michael Ekstrand", "links": { "alternate": { "href": "https://www.zotero.org/ekstrand", "type": "text/html" } } }, "creatorSummary": "Meng et al.", "parsedDate": "2020-09-22", "numChildren": 0 }, "data": { "key": "TNZ42I7U", "version": 230, "itemType": "conferencePaper", "title": "Exploring data splitting strategies for the evaluation of recommendation models", "creators": [ { "creatorType": "author", "firstName": "Zaiqiao", "lastName": "Meng" }, { "creatorType": "author", "firstName": "Richard", "lastName": "McCreadie" }, { "creatorType": "author", "firstName": "Craig", "lastName": "Macdonald" }, { "creatorType": "author", "firstName": "Iadh", "lastName": "Ounis" } ], "abstractNote": "Effective methodologies for evaluating recommender systems are critical,\nso that different systems can be compared in a sound manner. A commonly\noverlooked aspect of evaluating recommender systems is the selection of\nthe data splitting strategy. In this paper, we both show that there is no\nstandard splitting strategy and that the selection of splitting strategy\ncan have a strong impact on the ranking of recommender systems during\nevaluation. In particular, we perform experiments comparing three common\ndata splitting strategies, examining their impact over seven\nstate-of-the-art recommendation models on two datasets. Our results\ndemonstrate that the splitting strategy employed is an important\nconfounding variable that can markedly alter the ranking of recommender\nsystems, making much of the currently published literature non-comparable,\neven when the same datasets and metrics are used.", "proceedingsTitle": "Fourteenth ACM Conference on Recommender Systems", "conferenceName": "", "publisher": "Association for Computing Machinery", "place": "New York, NY, USA", "date": "2020-09-22", "eventPlace": "", "volume": "", "issue": "", "numberOfVolumes": "", "pages": "681-686", "series": "", "seriesNumber": "", "DOI": "10.1145/3383313.3418479", "ISBN": "978-1-4503-7583-2", "citationKey": "mengExploringDataSplitting2020", "url": "https://doi.org/10.1145/3383313.3418479", "accessDate": "2022-03-17", "ISSN": "", "archive": "", "archiveLocation": "", "shortTitle": "", "language": "", "libraryCatalog": "", "callNumber": "", "rights": "", "extra": "", "tags": [], "collections": [ "LSCSUCXH" ], "relations": {}, "dateAdded": "2024-12-21T21:10:23Z", "dateModified": "2026-02-16T15:58:04Z" } } ]