Cold start (recommender systems)

Last updated

Cold start is a potential problem in computer-based information systems which involves a degree of automated data modelling. Specifically, it concerns the issue that the system cannot draw any inferences for users or items about which it has not yet gathered sufficient information.

Contents

Systems affected

The cold start problem is a well known and well researched problem for recommender systems. Recommender systems form a specific type of information filtering (IF) technique that attempts to present information items (e-commerce, films, music, books, news, images, web pages) that are likely of interest to the user. Typically, a recommender system compares the user's profile to some reference characteristics. These characteristics may be related to item characteristics (content-based filtering) or the user's social environment and past behavior (collaborative filtering). Depending on the system, the user can be associated to various kinds of interactions: ratings, bookmarks, purchases, likes, number of page visits etc.

There are three cases of cold start: [1]

  1. New community: refers to the start-up of the recommender, when, although a catalogue of items might exist, almost no users are present and the lack of user interaction makes it very hard to provide reliable recommendations
  2. New item: a new item is added to the system, it might have some content information but no interactions are present
  3. New user: a new user registers and has not provided any interaction yet, therefore it is not possible to provide personalized recommendations

New community

The new community problem, or systemic bootstrapping, refers to the startup of the system, when virtually no information the recommender can rely upon is present. [2] This case presents the disadvantages of both the New user and the New item case, as all items and users are new. Due to this some of the techniques developed to deal with those two cases are not applicable to the system bootstrapping.

New item

The item cold-start problem refers to when items added to the catalogue have either none or very little interactions. This constitutes a problem mainly for collaborative filtering algorithms due to the fact that they rely on the item's interactions to make recommendations. If no interactions are available then a pure collaborative algorithm cannot recommend the item. In case only a few interactions are available, although a collaborative algorithm will be able to recommend it, the quality of those recommendations will be poor. [3] This raises another issue, which is not anymore related to new items, but rather to unpopular items. In some cases (e.g. movie recommendations) it might happen that a handful of items receive an extremely high number of interactions, while most of the items only receive a fraction of them. This is referred to as popularity bias. [4]

Number of user interactions associated to each item in a Movielens dataset. Few items have a very high number of interactions, more than 5000, while most of the others have less than 100 Popularity distribution Movielens dataset.png
Number of user interactions associated to each item in a Movielens dataset. Few items have a very high number of interactions, more than 5000, while most of the others have less than 100

In the context of cold-start items the popularity bias is important because it might happen that many items, even if they have been in the catalogue for months, received only a few interactions. This creates a negative loop in which unpopular items will be poorly recommended, therefore will receive much less visibility than popular ones, and will struggle to receive interactions. [5] While it is expected that some items will be less popular than others, this issue specifically refers to the fact that the recommender has not enough collaborative information to recommend them in a meaningful and reliable way. [6]

Content-based filtering algorithms, on the other hand, are in theory much less prone to the new item problem. Since content based recommenders choose which items to recommend based on the feature the items possess, even if no interaction for a new item exist, still its features will allow for a recommendation to be made. [7] This of course assumes that a new item will be already described by its attributes, which is not always the case. Consider the case of so-called editorial features (e.g. director, cast, title, year), those are always known when the item, in this case movie, is added to the catalogue. However, other kinds of attributes might not be e.g. features extracted from user reviews and tags. [8] Content-based algorithms relying on user provided features suffer from the cold-start item problem as well, since for new items if no (or very few) interactions exist, also no (or very few) user reviews and tags will be available.

New user

The new user case refers to when a new user enrolls in the system and for a certain period of time the recommender has to provide recommendation without relying on the user's past interactions, since none has occurred yet. [1] This problem is of particular importance when the recommender is part of the service offered to users, since a user who is faced with recommendations of poor quality might soon decide to stop using the system before providing enough interaction to allow the recommender to understand his/her interests. The main strategy in dealing with new users is to ask them to provide some preferences to build an initial user profile. A threshold has to be found between the length of the user registration process, which if too long might indice too many users to abandon it, and the amount of initial data required for the recommender to work properly. [2]

Similarly to the new items case, not all recommender algorithms are affected in the same way. Item-item recommenders will be affected as they rely on user profile to weight how relevant other user's preferences are. Collaborative filtering algorithms are the most affected as without interactions no inference can be made about the user's preferences. User-user recommender algorithms [9] behave slightly differently. A user-user content based algorithm will rely on user's features (e.g. age, gender, country) to find similar users and recommend the items they interacted with in a positive way, therefore being robust to the new user case. Note that all these information is acquired during the registration process, either by asking the user to input the data himself, or by leveraging data already available e.g. in his social media accounts. [10]

Mitigation strategies

Due to the high number of recommender algorithms available as well as system type and characteristics, many strategies to mitigate the cold-start problem have been developed. The main approach is to rely on hybrid recommenders, in order to mitigate the disadvantages of one category or model by combining it with another. [11] [12] [13]

All three categories of cold-start (new community, new item, and new user) have in common the lack of user interactions and presents some commonalities in the strategies available to address them.

A common strategy when dealing with new items is to couple a collaborative filtering recommender, for warm items, with a content-based filtering recommender, for cold-items. While the two algorithms can be combined in different ways, the main drawback of this method is related to the poor recommendation quality often exhibited by content-based recommenders in scenarios where it is difficult to provide a comprehensive description of the item characteristics. [14] In case of new users, if no demographic feature is present or their quality is too poor, a common strategy is to offer them non-personalized recommendations. This means that they could be recommended simply the most popular items either globally or for their specific geographical region or language.

Profile completion

One of the available options when dealing with cold users or items is to rapidly acquire some preference data. There are various ways to do that depending on the amount of information required. These techniques are called preference elicitation strategies. [15] [16] This may be done either explicitly (by querying the user) or implicitly (by observing the user's behaviour). In both cases, the cold start problem would imply that the user has to dedicate an amount of effort using the system in its 'dumb' state – contributing to the construction of their user profile – before the system can start providing any intelligent recommendations. [17]

For example MovieLens, a web-based recommender system for movies, asks the user to rate some movies as a part of the registration. While preference elicitation strategy are a simple and effective way to deal with new users, the additional requirements during the registration will make the process more time-consuming for the user. Moreover, the quality of the obtained preferences might not be ideal as the user could rate items they had seen months or years ago or the provided ratings could be almost random if the user provided them without paying attention just to complete the registration quickly.

The construction of the user's profile may also be automated by integrating information from other user activities, such as browsing histories or social media platforms. If, for example, a user has been reading information about a particular music artist from a media portal, then the associated recommender system would automatically propose that artist's releases when the user visits the music store. [18]

A variation of the previous approach is to automatically assign ratings to new items, based on the ratings assigned by the community to other similar items. Item similarity would be determined according to the items' content-based characteristics. [17]

It is also possible to create initial profile of a user based on the personality characteristics of the user and use such profile to generate personalized recommendation. [19] [20] Personality characteristics of the user can be identified using a personality model such as five factor model (FFM).

Another of the possible techniques is to apply active learning (machine learning). The main goal of active learning is to guide the user in the preference elicitation process in order to ask him to rate only the items that for the recommender point of view will be the most informative ones. This is done by analysing the available data and estimating the usefulness of the data points (e.g., ratings, interactions). [21] As an example, say that we want to build two clusters from a certain cloud of points. As soon as we have identified two points each belonging to a different cluster, which is the next most informative point? If we take a point close to one we already know we can expect that it will likely belong to the same cluster. If we choose a point which is in between the two clusters, knowing which cluster it belongs to will help us in finding where the boundary is, allowing to classify many other points with just a few observations.

The cold start problem is also exhibited by interface agents. Since such an agent typically learns the user's preferences implicitly by observing patterns in the user's behaviour – "watching over the shoulder" – it would take time before the agent may perform any adaptations personalised to the user. Even then, its assistance would be limited to activities which it has formerly observed the user engaging in. [22] The cold start problem may be overcome by introducing an element of collaboration amongst agents assisting various users. This way, novel situations may be handled by requesting other agents to share what they have already learnt from their respective users. [22]

Feature mapping

In recent years more advanced strategies have been proposed, they all rely on machine learning and attempt to merge the content and collaborative information in a single model. One example of this approaches is called attribute to feature mapping [23] which is tailored to matrix factorization algorithms. [24] The basic idea is the following. A matrix factorization model represents the user-item interactions as the product of two rectangular matrices whose content is learned using the known interactions via machine learning. Each user will be associated to a row of the first matrix and each item with a column of the second matrix. The row or column associated to a specific user or item is called latent factors. [25] When a new item is added it has no associated latent factors and the lack of interactions does not allow to learn them, as it was done with other items. If each item is associated to some features (e.g. author, year, publisher, actors) it is possible to define an embedding function, which given the item features estimates the corresponding item latent factors. The embedding function can be designed in many ways and it is trained with the data already available from warm items. Alternatively, one could apply a group-specific method. [26] [27] A group-specific method further decomposes each latent factor into two additive parts: One part corresponds to each item (and/or each user), while the other part is shared among items within each item group (e.g., a group of movies could be movies of the same genre). Then once a new item arrives, we can assign a group label to it, and approximates its latent factor by the group-specific part (of the corresponding item group). Therefore, although the individual part of the new item is not available, the group-specific part provides an immediate and effective solution. The same applies for a new user, as if some information is available for them (e.g. age, nationality, gender) then his/her latent factors can be estimated via an embedding function or a group-specific latent factor.

Hybrid feature weighting

Another recent approach which bears similarities with feature mapping is building a hybrid content-based filtering recommender in which features, either of the items or of the users, are weighted according to the user's perception of importance. In order to identify a movie that the user could like, different attributes (e.g. which are the actors, director, country, title) will have different importance. As an example consider the James Bond movie series, the main actor changed many times during the years, while some did not, like Lois Maxwell. Therefore, her presence will probably be a better identifier of that kind of movie than the presence of one of the various main actors. [14] [28] Although various techniques exist to apply feature weighting to user or item features in recommender systems, most of them are from the information retrieval domain like tf–idf, Okapi BM25, only a few have been developed specifically for recommenders. [29]

Hybrid feature weighting techniques in particular are tailored for the recommender system domain. Some of them learn feature weight by exploiting directly the user's interactions with items, like FBSM. [28] Others rely on an intermediate collaborative model trained on warm items and attempt to learn the content feature weights which will better approximate the collaborative model. [14]

Many of the hybrid methods can be considered special cases of factorization machines. [30] [31]

Differentiating regularization weights

The above methods rely on affiliated information from users or items. Recently, another approach mitigates the cold start problem by assigning lower constraints to the latent factors associated with the items or users that reveal more information (i.e., popular items and active users), and set higher constraints to the others (i.e., less popular items and inactive users). [32] It is shown that various recommendation models benefit from this strategy. Differentiating regularization weights can be integrated with the other cold start mitigating strategies.

See also

Related Research Articles

<span class="mw-page-title-main">Collaborative filtering</span> Algorithm

Collaborative filtering (CF) is a technique used by recommender systems. Collaborative filtering has two senses, a narrow one and a more general one.

A recommender system, or a recommendation system, is a subclass of information filtering system that provides suggestions for items that are most pertinent to a particular user. Recommender systems are particularly useful when an individual needs to choose an item from a potentially overwhelming number of items that a service may offer.

Preference elicitation refers to the problem of developing a decision support system capable of generating recommendations to a user, thus assisting in decision making. It is important for such a system to model user's preferences accurately, find hidden preferences and avoid redundancy. This problem is sometimes studied as a computational learning theory problem. Another approach for formulating this problem is a partially observable Markov decision process. The formulation of this problem is also dependent upon the context of the area in which it is studied.

Reputation systems are programs or algorithms that allow users to rate each other in online communities in order to build trust through reputation. Some common uses of these systems can be found on E-commerce websites such as eBay, Amazon.com, and Etsy as well as online advice communities such as Stack Exchange. These reputation systems represent a significant trend in "decision support for Internet mediated service provisions". With the popularity of online communities for shopping, advice, and exchange of other important information, reputation systems are becoming vitally important to the online experience. The idea of reputation systems is that even if the consumer can't physically try a product or service, or see the person providing information, that they can be confident in the outcome of the exchange through trust built by recommender systems.

User modeling is the subdivision of human–computer interaction which describes the process of building up and modifying a conceptual understanding of the user. The main goal of user modeling is customization and adaptation of systems to the user's specific needs. The system needs to "say the 'right' thing at the 'right' time in the 'right' way". To do so it needs an internal representation of the user. Another common purpose is modeling specific kinds of users, including modeling of their skills and declarative knowledge, for use in automatic software-tests. User-models can thus serve as a cheaper alternative to user testing but should not replace user testing.

Social information processing is "an activity through which collective human actions organize knowledge." It is the creation and processing of information by a group of people. As an academic field Social Information Processing studies the information processing power of networked social systems.

<span class="mw-page-title-main">GroupLens Research</span> Computer science research lab

GroupLens Research is a human–computer interaction research lab in the Department of Computer Science and Engineering at the University of Minnesota, Twin Cities specializing in recommender systems and online communities. GroupLens also works with mobile and ubiquitous technologies, digital libraries, and local geographic information systems.

Collaborative search engines (CSE) are Web search engines and enterprise searches within company intranets that let users combine their efforts in information retrieval (IR) activities, share information resources collaboratively using knowledge tags, and allow experts to guide less experienced people through their searches. Collaboration partners do so by providing query terms, collective tagging, adding comments or opinions, rating search results, and links clicked of former (successful) IR activities to users having the same or a related information need.

Collaborative tagging, also known as social tagging or folksonomy, allows users to apply public tags to online items, typically to make those items easier for themselves or others to find later. It has been argued that these tagging systems can provide navigational cues or "way-finders" for other users to explore information. The notion is that given that social tags are labels users create to represent topics extracted from online documents, the interpretation of these tags should allow other users to predict the contents of different documents efficiently. Social tags are arguably more important in exploratory search, in which the users may engage in iterative cycles of goal refinement and exploration of new information, and interpretation of information contents by others will provide useful cues for people to discover topics that are relevant.

MovieLens is a web-based recommender system and virtual community that recommends movies for its users to watch, based on their film preferences using collaborative filtering of members' movie ratings and movie reviews. It contains about 11 million ratings for about 8500 movies. MovieLens was created in 1997 by GroupLens Research, a research lab in the Department of Computer Science and Engineering at the University of Minnesota, in order to gather research data on personalized recommendations.

A content discovery platform is an implemented software recommendation platform which uses recommender system tools. It utilizes user metadata in order to discover and recommend appropriate content, whilst reducing ongoing maintenance and development costs. A content discovery platform delivers personalized content to websites, mobile devices and set-top boxes. A large range of content discovery platforms currently exist for various forms of content ranging from news articles and academic journal articles to television. As operators compete to be the gateway to home entertainment, personalized television is a key service differentiator. Academic content discovery has recently become another area of interest, with several companies being established to help academic researchers keep up to date with relevant academic content and serendipitously discover new content.

Personalized search is a web search tailored specifically to an individual's interests by incorporating information about the individual beyond the specific query provided. There are two general approaches to personalizing search results, involving modifying the user's query and re-ranking search results.

Gravity R&D is an IT vendor specialized in recommender systems. Gravity was founded by members of the Netflix Prize team "Gravity".

<span class="mw-page-title-main">User profile</span> Data about an individual user

A user profile is a collection of settings and information associated with a user. It contains critical information that is used to identify an individual, such as their name, age, portrait photograph and individual characteristics such as knowledge or expertise. User profiles are most commonly present on social media websites such as Facebook, Instagram, and LinkedIn; and serve as voluntary digital identity of an individual, highlighting their key features and traits. In personal computing and operating systems, user profiles serve to categorise files, settings, and documents by individual user environments, known as ‘accounts’, allowing the operating system to be more friendly and catered to the user. Physical user profiles serve as identity documents such as passports, driving licenses and legal documents that are used to identify an individual under the legal system.

<span class="mw-page-title-main">John T. Riedl</span> American computer scientist

John Thomas Riedl was an American computer scientist and the McKnight Distinguished Professor at the University of Minnesota. His published works include highly influential research on the social web, recommendation systems, and collaborative systems.

Knowledge-based recommender systems are a specific type of recommender system that are based on explicit knowledge about the item assortment, user preferences, and recommendation criteria. These systems are applied in scenarios where alternative approaches such as collaborative filtering and content-based filtering cannot be applied.

Item-item collaborative filtering, or item-based, or item-to-item, is a form of collaborative filtering for recommender systems based on the similarity between items calculated using people's ratings of those items. Item-item collaborative filtering was invented and used by Amazon.com in 1998. It was first published in an academic conference in 2001.

Social navigation is a form of social computing introduced by Paul Dourish and Matthew Chalmers in 1994, who defined it as when "movement from one item to another is provoked as an artifact of the activity of another or a group of others". According to later research in 2002, "social navigation exploits the knowledge and experience of peer users of information resources" to guide users in the information space, and that it is becoming more difficult to navigate and search efficiently with all the digital information available from the World Wide Web and other sources. Studying others' navigational trails and understanding their behavior can help improve one's own search strategy by guiding them to make more informed decisions based on the actions of others.

Video browsing, also known as exploratory video search, is the interactive process of skimming through video content in order to satisfy some information need or to interactively check if the video content is relevant. While originally proposed to help users inspecting a single video through visual thumbnails, modern video browsing tools enable users to quickly find desired information in a video archive by iterative human–computer interaction through an exploratory search approach. Many of these tools presume a smart user that wants features to interactively inspect video content, as well as automatic content filtering features. For that purpose, several video interaction features are usually provided, such as sophisticated navigation in video or search by a content-based query. Video browsing tools often build on lower-level video content analysis, such as shot transition detection, keyframe extraction, semantic concept detection, and create a structured content overview of the video file or video archive. Furthermore, they usually provide sophisticated navigation features, such as advanced timelines, visual seeker bars or a list of selected thumbnails, as well as means for content querying. Examples of content queries are shot filtering through visual concepts, through some specific characteristics, through user-provided sketches, or through content-based similarity search.

Matrix factorization is a class of collaborative filtering algorithms used in recommender systems. Matrix factorization algorithms work by decomposing the user-item interaction matrix into the product of two lower dimensionality rectangular matrices. This family of methods became widely known during the Netflix prize challenge due to its effectiveness as reported by Simon Funk in his 2006 blog post, where he shared his findings with the research community. The prediction results can be improved by assigning different regularization weights to the latent factors based on items' popularity and users' activeness.

References

  1. 1 2 Bobadilla, Jesús; Ortega, Fernando; Hernando, Antonio; Bernal, Jesús (February 2012). "A collaborative filtering approach to mitigate the new user cold start problem". Knowledge-Based Systems. 26: 225–238. doi:10.1016/j.knosys.2011.07.021.
  2. 1 2 Rashid, Al Mamunur; Karypis, George; Riedl, John (20 December 2008). "Learning preferences of new users in recommender systems". ACM SIGKDD Explorations Newsletter. 10 (2): 90. doi:10.1145/1540276.1540302. S2CID   1651200.
  3. ↑ Lika, Blerina; Kolomvatsos, Kostas; Hadjiefthymiades, Stathes (March 2014). "Facing the cold start problem in recommender systems". Expert Systems with Applications. 41 (4): 2065–2073. doi:10.1016/j.eswa.2013.09.005. S2CID   3491063.
  4. ↑ Hou, Lei; Pan, Xue; Liu, Kecheng (7 March 2018). "Balancing the popularity bias of object similarities for personalised recommendation". The European Physical Journal B. 91 (3): 47. Bibcode:2018EPJB...91...47H. doi: 10.1140/epjb/e2018-80374-8 .
  5. ↑ Abdollahpouri, Himan; Burke, Robin; Mobasher, Bamshad (27 August 2017). Proceedings of the Eleventh ACM Conference on Recommender Systems - Rec Sys '17. ACM. pp. 42–46. doi:10.1145/3109859.3109912. ISBN   9781450346528. S2CID   10104149.
  6. ↑ Park, Yoon-Joo; Tuzhilin, Alexander (23 October 2008). Proceedings of the 2008 ACM conference on Recommender systems - Rec Sys '08. ACM. pp. 11–18. CiteSeerX   10.1.1.421.1833 . doi:10.1145/1454008.1454012. ISBN   9781605580937. S2CID   6833818.
  7. ↑ Pazzani, Michael J.; Billsus, Daniel (2007). "Content-Based Recommendation Systems". The Adaptive Web. Lecture Notes in Computer Science. Vol. 4321. pp. 325–341. CiteSeerX   10.1.1.130.8327 . doi:10.1007/978-3-540-72079-9_10. ISBN   978-3-540-72078-2. S2CID   7364032.
  8. ↑ Chen, Li; Chen, Guanliang; Wang, Feng (22 January 2015). "Recommender systems based on user reviews: the state of the art". User Modeling and User-Adapted Interaction. 25 (2): 99–154. doi:10.1007/s11257-015-9155-5. S2CID   7847519.
  9. ↑ Bobadilla, J.; Ortega, F.; Hernando, A.; Gutiérrez, A. (July 2013). "Recommender systems survey". Knowledge-Based Systems. 46: 109–132. doi:10.1016/j.knosys.2013.03.012. S2CID   206710454.
  10. ↑ Zhang, Zi-Ke; Liu, Chuang; Zhang, Yi-Cheng; Zhou, Tao (1 October 2010). "Solving the cold-start problem in recommender systems with social tags". EPL (Europhysics Letters). 92 (2): 28002. arXiv: 1004.3732 . Bibcode:2010EL.....9228002Z. doi:10.1209/0295-5075/92/28002.
  11. ↑ Huang, Zan; Chen, Hsinchun; Zeng, Daniel (1 January 2004). "Applying associative retrieval techniques to alleviate the sparsity problem in collaborative filtering". ACM Transactions on Information Systems. 22 (1): 116–142. CiteSeerX   10.1.1.3.1590 . doi:10.1145/963770.963775. S2CID   13561350.
  12. ↑ Salter, J.; Antonopoulos, N. (January 2006). "CinemaScreen Recommender Agent: Combining Collaborative and Content-Based Filtering" (PDF). IEEE Intelligent Systems. 21 (1): 35–41. doi:10.1109/MIS.2006.4. S2CID   15715899.
  13. ↑ Burke, Robin (2007). "Hybrid Web Recommender Systems". The Adaptive Web. Lecture Notes in Computer Science. Vol. 4321. pp. 377–408. CiteSeerX   10.1.1.395.8975 . doi:10.1007/978-3-540-72079-9_12. ISBN   978-3-540-72078-2. S2CID   183823.
  14. 1 2 3 Cella, Leonardo; Cereda, Stefano; Quadrana, Massimo; Cremonesi, Paolo (2017). "Deriving Item Features Relevance from Past User Interactions". Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization. pp. 275–279. doi:10.1145/3079628.3079695. hdl:11311/1061220. ISBN   9781450346351. S2CID   11459891.
  15. ↑ Elahi, Mehdi; Ricci, Francesco; Rubens, Neil (2014). E-Commerce and Web Technologies. Lecture Notes in Business Information Processing. Vol. 188. Springer International Publishing. pp. 113–124. doi:10.1007/978-3-319-10491-1_12. ISBN   978-3-319-10491-1.
  16. ↑ Elahi, Mehdi; Ricci, Francesco; Rubens, Neil (2016). "A survey of active learning in collaborative filtering recommender systems". Computer Science Review. 20: 29–50. doi:10.1016/j.cosrev.2016.05.002 – via Elsevier.
  17. 1 2 Andrew I. Schein; Alexandrin Popescul; Lyle H. Ungar; David M. Pennock (2002). Methods and Metrics for Cold-Start Recommendations . Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2002). New York City, New York: ACM. pp.  253–260. ISBN   1-58113-561-0 . Retrieved 2008-02-02.
  18. ↑ "Vendor attempts to crack 'cold start' problem in content recommendations" (PDF). Mobile Media: 18. 2007-06-29. Archived from the original (PDF) on 2008-11-21. Retrieved 2008-02-02.
  19. ↑ Tkalcic, Marko; Chen, Li (2016). "Personality and Recommender Systems". In Ricci, Francesco; Rokach, Lior; Shapira, Bracha (eds.). Recommender Systems Handbook (2nd ed.). Springer US. doi:10.1007/978-1-4899-7637-6_21. ISBN   978-1-4899-7637-6.
  20. ↑ Fernández-Tobías, Ignacio; Braunhofer, Matthias; Elahi, Mehdi; Ricci, Francesco; Cantador, Iván (2016). "Alleviating the new user problem in collaborative filtering by exploiting personality information". User Modeling and User-Adapted Interaction. 26 (2–3): 221–255. doi:10.1007/s11257-016-9172-z. hdl: 10486/674370 . S2CID   8687279.
  21. ↑ Rubens, Neil; Elahi, Mehdi; Sugiyama, Masashi; Kaplan, Dain (2016). "Active Learning in Recommender Systems". In Ricci, Francesco; Rokach, Lior; Shapira, Bracha (eds.). Recommender Systems Handbook (2nd ed.). Springer US. doi:10.1007/978-1-4899-7637-6_24. ISBN   978-1-4899-7637-6.
  22. 1 2 Yezdi Lashkari; Max Metral; Pattie Maes (1994). Collaborative Interface Agents. Proceedings of the Twelfth National Conference on Artificial Intelligence. Seattle, Washington: AAAI Press. pp. 444–449. ISBN   0-262-61102-3 . Retrieved 2008-02-02.
  23. ↑ Gantner, Zeno; Drumond, Lucas; Freudenthaler, Cristoph (20 January 2011). 2010 IEEE International Conference on Data Mining. pp. 176–185. CiteSeerX   10.1.1.187.5933 . doi:10.1109/ICDM.2010.129. ISBN   978-1-4244-9131-5. S2CID   8955774.
  24. ↑ Koren, Yehuda; Bell, Robert; Volinsky, Chris (August 2009). "Matrix Factorization Techniques for Recommender Systems". Computer. 42 (8): 30–37. CiteSeerX   10.1.1.147.8295 . doi:10.1109/MC.2009.263. S2CID   58370896.
  25. ↑ Agarwal, Deepak; Chen, Bee-Chung (28 June 2009). Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining - KDD '09. ACM. pp. 19–28. doi:10.1145/1557019.1557029. ISBN   9781605584959. S2CID   17484284.
  26. ↑ Bi, Xuan; Qu, Annie; Wang, Junhui; Shen, Xiaotong (2017). "A group-specific recommender system". Journal of the American Statistical Association. 112 (519): 1344–1353. doi:10.1080/01621459.2016.1219261. S2CID   125187672.
  27. ↑ Bi, Xuan; Qu, Annie; Shen, Xiaotong (2018). "Multilayer tensor factorization with applications to recommender systems". Annals of Statistics. 46 (6B): 3303–3333. arXiv: 1711.01598 . doi:10.1214/17-AOS1659. S2CID   13677707.
  28. 1 2 Sharma, Mohit; Zhou, Jiayu; Hu, Junling; Karypis, George (2015). "Feature-based factorized Bilinear Similarity Model for Cold-Start Top- n Item Recommendation". Proceedings of the 2015 SIAM International Conference on Data Mining. pp. 190–198. doi:10.1137/1.9781611974010.22. ISBN   978-1-61197-401-0. S2CID   15887870.
  29. ↑ Symeonidis, Panagiotis; Nanopoulos, Alexandros; Manolopoulos, Yannis (25 July 2007). "Feature-Weighted User Model for Recommender Systems". User Modeling 2007. Lecture Notes in Computer Science. Vol. 4511. pp. 97–106. doi:10.1007/978-3-540-73078-1_13. ISBN   978-3-540-73077-4.
  30. ↑ Rendle, Steffen (1 May 2012). "Factorization Machines with libFM". ACM Transactions on Intelligent Systems and Technology. 3 (3): 1–22. doi:10.1145/2168752.2168771. S2CID   5499886.
  31. ↑ Rendle, Steffen (2010). "Factorization Machines". 2010 IEEE International Conference on Data Mining. IEEE. pp. 995–1000. CiteSeerX   10.1.1.393.8529 . doi:10.1109/ICDM.2010.127. ISBN   9781424491315. S2CID   17265929.
  32. ↑ ChenHung-Hsuan; ChenPu (2019-01-09). "Differentiating Regularization Weights -- A Simple Mechanism to Alleviate Cold Start in Recommender Systems". ACM Transactions on Knowledge Discovery from Data. 13: 1–22. doi:10.1145/3285954. S2CID   59337456.
This page is based on this Wikipedia article
Text is available under the CC BY-SA 4.0 license; additional terms may apply.
Images, videos and audio are available under their respective licenses.