Dawen Liang, Jaan Altosaar, Laurent Charlin, D. Blei
Matrix factorization (MF) models and their extensions are standard in modern recommender systems. MF models decompose the observed user-item interaction matrix into user and item latent factors. In this paper, we propose a co-factorization model, CoFactor, which jointly decomposes the user-item interaction matrix and the item-item co-occurrence matrix with shared item latent factors. For each pair of items, the co-occurrence matrix encodes the number of users that have consumed both items. CoFactor is inspired by the recent success of word embedding models (e.g., word2vec) which can be interpreted as factorizing the word co-occurrence matrix. We show that this model significantly improves the performance over MF models on several datasets with little additional computational overhead. We provide qualitative results that explain how CoFactor improves the quality of the inferred factors and characterize the circumstances where it provides the most significant improvements.
{"title":"Factorization Meets the Item Embedding: Regularizing Matrix Factorization with Item Co-occurrence","authors":"Dawen Liang, Jaan Altosaar, Laurent Charlin, D. Blei","doi":"10.1145/2959100.2959182","DOIUrl":"https://doi.org/10.1145/2959100.2959182","url":null,"abstract":"Matrix factorization (MF) models and their extensions are standard in modern recommender systems. MF models decompose the observed user-item interaction matrix into user and item latent factors. In this paper, we propose a co-factorization model, CoFactor, which jointly decomposes the user-item interaction matrix and the item-item co-occurrence matrix with shared item latent factors. For each pair of items, the co-occurrence matrix encodes the number of users that have consumed both items. CoFactor is inspired by the recent success of word embedding models (e.g., word2vec) which can be interpreted as factorizing the word co-occurrence matrix. We show that this model significantly improves the performance over MF models on several datasets with little additional computational overhead. We provide qualitative results that explain how CoFactor improves the quality of the inferred factors and characterize the circumstances where it provides the most significant improvements.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128889781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Machine Learning research is progressing at an ever increasing pace. Fueled by technology advances commonly referred to as "Big Data", all data related fields are teaming with scientific and applied activity: our communities explore new application areas, develop new learning algorithms, and continuously scale and improve optimization and estimation methods. But from an industry perspective, many of the most impeding challenges are entirely elsewhere. This talk takes a fresh look at the practical state of affairs in the context of running a large-scale automated machine learning system that supports 50 Billion decision daily on behalf of hundreds of digital advertisers. Some of the key lessons are 1) robustness beats peak performance almost always, 2) support for the constant dynamic fluctuations in the data stream is essential, 3) models exploiting unknowingly any weakness of your metrics, and finally 4) the fact that despite big data, the data you really want never exists.
{"title":"Automated Machine Learning in the Wild","authors":"C. Perlich","doi":"10.1145/2959100.2959191","DOIUrl":"https://doi.org/10.1145/2959100.2959191","url":null,"abstract":"Machine Learning research is progressing at an ever increasing pace. Fueled by technology advances commonly referred to as \"Big Data\", all data related fields are teaming with scientific and applied activity: our communities explore new application areas, develop new learning algorithms, and continuously scale and improve optimization and estimation methods. But from an industry perspective, many of the most impeding challenges are entirely elsewhere. This talk takes a fresh look at the practical state of affairs in the context of running a large-scale automated machine learning system that supports 50 Billion decision daily on behalf of hundreds of digital advertisers. Some of the key lessons are 1) robustness beats peak performance almost always, 2) support for the constant dynamic fluctuations in the data stream is essential, 3) models exploiting unknowingly any weakness of your metrics, and finally 4) the fact that despite big data, the data you really want never exists.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130689594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Preparing recommendations for unknown users or such that correctly respond to the short-term needs of a particular user is one of the fundamental problems for e-commerce. Most of the common Recommender Systems assume that user identification must be explicit. In this paper a Session-Aware Recommender System approach is presented where no straightforward user information is required. The recommendation process is based only on user activity within a single session, defined as a sequence of events. This information is incorporated in the recommendation process by explicit context modeling with factorization methods and a novel approach with Recurrent Neural Network (RNN). Compared to the session modeling approach, RNN directly models the dependency of user observed sequential behavior throughout its recurrent structure. The evaluation discusses the results based on sessions from real-life system with ephemeral items (identified only by the set of their attributes) for the task of top-n best recommendations.
{"title":"Modelling Contextual Information in Session-Aware Recommender Systems with Neural Networks","authors":"Bartlomiej Twardowski","doi":"10.1145/2959100.2959162","DOIUrl":"https://doi.org/10.1145/2959100.2959162","url":null,"abstract":"Preparing recommendations for unknown users or such that correctly respond to the short-term needs of a particular user is one of the fundamental problems for e-commerce. Most of the common Recommender Systems assume that user identification must be explicit. In this paper a Session-Aware Recommender System approach is presented where no straightforward user information is required. The recommendation process is based only on user activity within a single session, defined as a sequence of events. This information is incorporated in the recommendation process by explicit context modeling with factorization methods and a novel approach with Recurrent Neural Network (RNN). Compared to the session modeling approach, RNN directly models the dependency of user observed sequential behavior throughout its recurrent structure. The evaluation discusses the results based on sessions from real-life system with ephemeral items (identified only by the set of their attributes) for the task of top-n best recommendations.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130867856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Were The Rolling Stones right when they said, "You can't always get what you want; but if you try sometime you get what you need"? Recommendation systems are the crystal ball of the Internet: predicting user intentions, making sense of big data, and delivering what people are looking for before they even know they want it. Pandora radio is best known for the Music Genome Project; the most unique and richly labeled music catalog of 1.5 million+ tracks. While this content-based approach to music recommendation is extremely effective and still used today as the foundation to the leading online radio service, Pandora has also collected more than a decade of contextual listener feedback in the form of more than 65 billion thumbs from 79M+ monthly active users who have created more than 9 billion stations. This session will look at how the interdisciplinary team at Pandora goes about making sense of these massive data sets to successfully make large scale music recommendations to our listeners. As opposed to more traditional recommender systems which need only to recommend a single item or set of items, Pandora's recommenders must provide an evolving set of sequential items, which constantly keep the experience new and exciting. In this talk I will present a dynamic ensemble learning system that combines musicological data and machine learning models to provide a truly personalized experience. This approach allows us to switch from a lean back experience (exploitation) to a more exploration mode to discover new music tailored specifically to users individual tastes. To exemplify this, I will present a recently launched product led by the research team, Thumbprint Radio. Following this session the audience will have an in-depth understanding of how Pandora uses science to determine the perfect balance of familiarity, discovery, repetition and relevance for each individual listener, measures and evaluates user satisfaction, and how our online and offline architecture stack plays a critical role in our success.
{"title":"The Exploit-Explore Dilemma in Music Recommendation","authors":"Òscar Celma","doi":"10.1145/2959100.2959122","DOIUrl":"https://doi.org/10.1145/2959100.2959122","url":null,"abstract":"Were The Rolling Stones right when they said, \"You can't always get what you want; but if you try sometime you get what you need\"? Recommendation systems are the crystal ball of the Internet: predicting user intentions, making sense of big data, and delivering what people are looking for before they even know they want it. Pandora radio is best known for the Music Genome Project; the most unique and richly labeled music catalog of 1.5 million+ tracks. While this content-based approach to music recommendation is extremely effective and still used today as the foundation to the leading online radio service, Pandora has also collected more than a decade of contextual listener feedback in the form of more than 65 billion thumbs from 79M+ monthly active users who have created more than 9 billion stations. This session will look at how the interdisciplinary team at Pandora goes about making sense of these massive data sets to successfully make large scale music recommendations to our listeners. As opposed to more traditional recommender systems which need only to recommend a single item or set of items, Pandora's recommenders must provide an evolving set of sequential items, which constantly keep the experience new and exciting. In this talk I will present a dynamic ensemble learning system that combines musicological data and machine learning models to provide a truly personalized experience. This approach allows us to switch from a lean back experience (exploitation) to a more exploration mode to discover new music tailored specifically to users individual tastes. To exemplify this, I will present a recently launched product led by the research team, Thumbprint Radio. Following this session the audience will have an in-depth understanding of how Pandora uses science to determine the perfect balance of familiarity, discovery, repetition and relevance for each individual listener, measures and evaluates user satisfaction, and how our online and offline architecture stack plays a critical role in our success.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123254375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
David Elsweiler, Bernd Ludwig, A. Said, Hanna Schäfer, C. Trattner
The first Workshop on Engendering Health with Recommender Systems was organized in conjunction with ACM RecSys 2016. The focus of the workshop was on bringing together researchers and practitioners from diverse areas of health, well-being, decision support, and behavioral change. Health-related issues in recommender systems have been a growing research topic in the recent years and this was a initial attempt at bringing together academics and practitioners to share their experiences on working on related issues.
{"title":"Engendering Health with Recommender Systems","authors":"David Elsweiler, Bernd Ludwig, A. Said, Hanna Schäfer, C. Trattner","doi":"10.1145/2959100.2959203","DOIUrl":"https://doi.org/10.1145/2959100.2959203","url":null,"abstract":"The first Workshop on Engendering Health with Recommender Systems was organized in conjunction with ACM RecSys 2016. The focus of the workshop was on bringing together researchers and practitioners from diverse areas of health, well-being, decision support, and behavioral change. Health-related issues in recommender systems have been a growing research topic in the recent years and this was a initial attempt at bringing together academics and practitioners to share their experiences on working on related issues.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123343771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Balázs Hidasi, Massimo Quadrana, Alexandros Karatzoglou, D. Tikk
Real-life recommender systems often face the daunting task of providing recommendations based only on the clicks of a user session. Methods that rely on user profiles -- such as matrix factorization -- perform very poorly in this setting, thus item-to-item recommendations are used most of the time. However the items typically have rich feature representations such as pictures and text descriptions that can be used to model the sessions. Here we investigate how these features can be exploited in Recurrent Neural Network based session models using deep learning. We show that obvious approaches do not leverage these data sources. We thus introduce a number of parallel RNN (p-RNN) architectures to model sessions based on the clicks and the features (images and text) of the clicked items. We also propose alternative training strategies for p-RNNs that suit them better than standard training. We show that p-RNN architectures with proper training have significant performance improvements over feature-less session models while all session-based models outperform the item-to-item type baseline.
{"title":"Parallel Recurrent Neural Network Architectures for Feature-rich Session-based Recommendations","authors":"Balázs Hidasi, Massimo Quadrana, Alexandros Karatzoglou, D. Tikk","doi":"10.1145/2959100.2959167","DOIUrl":"https://doi.org/10.1145/2959100.2959167","url":null,"abstract":"Real-life recommender systems often face the daunting task of providing recommendations based only on the clicks of a user session. Methods that rely on user profiles -- such as matrix factorization -- perform very poorly in this setting, thus item-to-item recommendations are used most of the time. However the items typically have rich feature representations such as pictures and text descriptions that can be used to model the sessions. Here we investigate how these features can be exploited in Recurrent Neural Network based session models using deep learning. We show that obvious approaches do not leverage these data sources. We thus introduce a number of parallel RNN (p-RNN) architectures to model sessions based on the clicks and the features (images and text) of the clicked items. We also propose alternative training strategies for p-RNNs that suit them better than standard training. We show that p-RNN architectures with proper training have significant performance improvements over feature-less session models while all session-based models outperform the item-to-item type baseline.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"92 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123093750","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
One of the strong points of E-commerce websites is that they are often abundant with product reviews from consumers who experienced the products and testify to the usefulness of the products or otherwise. These reviews are helpful for consumers to optimize their purchasing decisions. However, while popular products receive many reviews, many other products do not have an adequate number of reviews leading to the cold item problem. In this proposal, we propose a solution outline for the cold item problem by automatically generating reviews and predicting ratings for the cold products from available reviews of similar products in e-commerce websites as well as users' opinion shared in the microblogging platforms such as Twitter. We propose a framework to build a formal semantic representation of products from unstructured product descriptions, user reviews as well as user ratings. Such presentations assist us to measure product similarity and relatedness in a accurate and cost-effective way. Besides, we propose a model to generate additional reviews for a cold product by mining users' posts shared on medium such as Twitter and transfer them to the e-commerce website. Preliminary experiments show promising results in finding products similar to the cold products.
{"title":"Mining Information for the Cold-Item Problem","authors":"F. Pourgholamali","doi":"10.1145/2959100.2959102","DOIUrl":"https://doi.org/10.1145/2959100.2959102","url":null,"abstract":"One of the strong points of E-commerce websites is that they are often abundant with product reviews from consumers who experienced the products and testify to the usefulness of the products or otherwise. These reviews are helpful for consumers to optimize their purchasing decisions. However, while popular products receive many reviews, many other products do not have an adequate number of reviews leading to the cold item problem. In this proposal, we propose a solution outline for the cold item problem by automatically generating reviews and predicting ratings for the cold products from available reviews of similar products in e-commerce websites as well as users' opinion shared in the microblogging platforms such as Twitter. We propose a framework to build a formal semantic representation of products from unstructured product descriptions, user reviews as well as user ratings. Such presentations assist us to measure product similarity and relatedness in a accurate and cost-effective way. Besides, we propose a model to generate additional reviews for a cold product by mining users' posts shared on medium such as Twitter and transfer them to the e-commerce website. Preliminary experiments show promising results in finding products similar to the cold products.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114173821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We believe that in the future, the most common form of recommender systems will be present in a personal assistant. We claim that such an intelligent agent must be personal, i.e., know its user's preferences and recommend relevant content, a dynamic learner, instructable, supportive and affable. We describe the current state of the art and the challenges which should be addressed in each of these agent properties and provide examples of how we expect future personal agents to convey these properties.
{"title":"Recommender Systems with Personality","authors":"A. Azaria, Jason I. Hong","doi":"10.1145/2959100.2959138","DOIUrl":"https://doi.org/10.1145/2959100.2959138","url":null,"abstract":"We believe that in the future, the most common form of recommender systems will be present in a personal assistant. We claim that such an intelligent agent must be personal, i.e., know its user's preferences and recommend relevant content, a dynamic learner, instructable, supportive and affable. We describe the current state of the art and the challenges which should be addressed in each of these agent properties and provide examples of how we expect future personal agents to convey these properties.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"342 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122757085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The 4th Workshop on Emotions and Personality in Personalized Systems (EMPIRE) is taking place in Boston on September 16th, 2016 in conjunction with the ACM RecSys 2016 conference. The workshop focuses on the acquisition and usage of emotions and personality as user-centric aspects of personalization.
{"title":"4th Workshop on Emotions and Personality in Personalized Systems (EMPIRE)","authors":"M. Tkalcic, B. D. Carolis, M. Degemmis, A. Košir","doi":"10.1145/2959100.2959201","DOIUrl":"https://doi.org/10.1145/2959100.2959201","url":null,"abstract":"The 4th Workshop on Emotions and Personality in Personalized Systems (EMPIRE) is taking place in Boston on September 16th, 2016 in conjunction with the ACM RecSys 2016 conference. The workshop focuses on the acquisition and usage of emotions and personality as user-centric aspects of personalization.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134134528","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the ubiquity of fast internet connections and the growing availability of Video-On-Demand (VOD) services powerful recommender systems are needed. Traditionally, movie recommender systems apply user-based collaborative filtering providing high quality recommendations if users maintain user profiles describing preferences and movie ratings. The shortcomings of Collaborative Filtering are that comprehensive user profiles are required and users tend to get recommendations very similar to the user profile "filter bubble". In addition, CF-based recommenders neither consider current trends nor the context. In order to overcome these weaknesses, we develop a system identifying interesting events in the stream of current news and deploying this information for computing recommendations. Our system gathers topics of interest from Twitter and RSS-Feeds, extracts relevant Named Entities, and uses semantic relations for recommending movies closely related to these topics. We explain the used algorithms and show that our system provides highly relevant recommendations.
{"title":"Topical Semantic Recommendations for Auteur Films","authors":"Christian Rakow, A. Lommatzsch, Till Plumbaum","doi":"10.1145/2959100.2959110","DOIUrl":"https://doi.org/10.1145/2959100.2959110","url":null,"abstract":"With the ubiquity of fast internet connections and the growing availability of Video-On-Demand (VOD) services powerful recommender systems are needed. Traditionally, movie recommender systems apply user-based collaborative filtering providing high quality recommendations if users maintain user profiles describing preferences and movie ratings. The shortcomings of Collaborative Filtering are that comprehensive user profiles are required and users tend to get recommendations very similar to the user profile \"filter bubble\". In addition, CF-based recommenders neither consider current trends nor the context. In order to overcome these weaknesses, we develop a system identifying interesting events in the stream of current news and deploying this information for computing recommendations. Our system gathers topics of interest from Twitter and RSS-Feeds, extracts relevant Named Entities, and uses semantic relations for recommending movies closely related to these topics. We explain the used algorithms and show that our system provides highly relevant recommendations.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134078130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：