Multiverse recommendation: n-dimensional tensor factorization for context-aware collaborative filtering
By Alexandros Karatzoglou, Xavier Amatriain, Linas Baltrunas, Nuria Oliver. 2010
860 citations recorded by the source corpus.
Context has been recognized as an important factor to consider in personalized Recommender Systems. However, most model-based Collaborative Filtering approaches such as Matrix Factorization do not provide a straightforward way of integrating context information into the model. In this work, we introduce a Collaborative Filtering method based on Tensor Factorization, a generalization of Matrix Factorization that allows for a flexible and generic integration of contextual information by modeling the data as a User-Item-Context N-dimensional tensor instead of the traditional 2D User-Item matrix. In the proposed model, called Multiverse Recommendation, different types of context are considered as additional dimensions in the representation of the data as a tensor. The factorization of this tensor leads to a compact model of the data which can be used to provide context-aware recommendations. We provide an algorithm to address the N-dimensional factorization, and show that the Multiverse Recommendation improves upon non-contextual Matrix Factorization up to 30% in terms of the Mean Absolute Error (MAE). We also compare to two state-of-the-art context-aware methods and show that Tensor Factorization consistently outperforms them both in semi-synthetic and real-world data - improvements range from 2.5% to more than 12% depending on the data. Noticeably, our approach outperforms other methods by a wider margin whenever more contextual information is available.
Subfields: Computational Mathematics, Information Systems, Management Science and Operations Research
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Mathematics
- Charu C. Aggarwal
- Steffen Rendle
- Yu Zheng
- N. Sidiropoulos
- Jimeng Sun
- Christos Faloutsos
- Evangelos E. Papalexakis
- Balázs Hidasi
In Information Systems
- Xiangnan He
- Yongfeng Zhang
- Xiang Wang
- Tat-seng Chua
- Charu C. Aggarwal
- Xu Chen
- Julian McAuley
- Steffen Rendle
In Management Science and Operations Research
- Xiangnan He
- Julian McAuley
- Steffen Rendle
- Charu C. Aggarwal
- Xiang Wang
- Alexandros Karatzoglou
- Balázs Hidasi
- Meng Wang
Citations in the corpus
- Improving stock market prediction via heterogeneous information fusion (2017, methodology): “Thus, we use an element-wise optimization algorithm to iteratively update each entry in the matrices and tensor independently by gradient descent [50, 51].”
- TFMAP: optimizing MAP for top-n context-aware recommendation (2012, methodology): “Two state-of-theart approaches have been proposed to date, one based on tensor factorization [14, 35] and the other on factorization machines [25] (FM).”
- A Survey on Accuracy-Oriented Neural Recommendation: From Collaborative Filtering to Information-Rich Recommendation (2021, background): “zation (MF) [3] later becomes the mainstream recommender model for a long time (from 2008 until 2016) [4], [5].”
- Tensors for Data Mining and Data Fusion (2016, methodology): “In a similar spirit as earlier and around the same time, Karatzoglou et al. [2010] propose to use context (such as time) in traditional user-item recommendation scenarios by modeling the data as a tensor.”
- Fast context-aware recommendations with factorization machines (2011, methodology): “Multiverse Recommendation is learned with an SGD algorithm similar to the one proposed in [5].”
- User Conditional Hashtag Prediction for Images (2015, background): “Many works have incorporated user modeling into machine learning methods for various applications, for example by considering tensor factorisations for recommendation using music tags [13], web pages [10], using age and gender [4], location [21] and time [19].”
- SUSTain: Scalable Unsupervised Scoring for Tensors and its Application to Phenotyping (2018, background): “They have been applied successfully in diverse applications, including social network analysis [31], text mining [14], image processing [35], recommendation systems [26], brain data analysis [2] and healthcare analytics [22], to name a few.”
- Meta-Prod2Vec: Product Embeddings Using Side-Information for Recommendation (2016, methodology): “Tensor factorization have been suggested as a generalization of MF for considering additional information [12].”
- Mining contextual movie similarity with matrix factorization for context-aware recommendation (2013, background): “Under such approaches, latent features are learned from the {user, tag, item}triplet/ternary data directly for item recommendation [Karatzoglou et al. 2010], tag recommendation [Rendle et al. 2009b; Rendle and Schmidt-Thieme 2010; Symeonidis et al. 2008], or both [Symeonidis et al. 2010].”
- Diagnosing New York city's noises with ubiquitous data (2014, methodology): “Specifically, we use an element-wise optimization algorithm [6], which updates each entry in the tensor independently.”
- Logistic Matrix Factorization for Implicit Feedback Data (2014, background): “Context-aware collaborative filtering attempts to take into account the context with which users provided implicit feedback, such as the time of day, day of the week, or location.”
- Tensor Factorization for Low-Rank Tensor Completion (2018, background): “So tensor analysis is of practical significance and benefits many applications in computer vision [2], [6], [7], data mining [8], collaborative filtering [9], etc.”
- Sparse linear methods with side information for top-n recommendations (2012, methodology): “Such approaches include hybrid methods [5], matrix/tensor factorization [14, 8], and other regression methods [1].”
- Improving pairwise learning for item recommendation from implicit feedback (2014, neutral): “[9, 13]).”