Maximum Margin Matrix Factorization for Collaborative Ranking

By Markus Weimer, Alexandros Karatzoglou, Quoc V. Le, A. Smola. 2007

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Xiangnan He
2. Yoshua Bengio
3. Dacheng Tao
4. Hanwang Zhang
5. Yi Tay
6. Xiang Wang
7. Trevor Darrell
8. Bernt Schiele

Citations in the corpus

• Improved Recurrent Neural Networks for Session-based Recommendations (2016, methodology): “Matrix factorization methods [15, 30] are based on the sparse user-item interaction matrix, where the recommendation problem is formulated as a matrix completion task.”
• Session-based Recommendations with Recurrent Neural Networks (2015, neutral): “The most common methods used in recommender systems are factor models (Koren et al., 2009; Weimer et al., 2007; Hidasi & Tikk, 2012) and neighborhood methods (Sarwar et al., 2001; Ko-ren, 2008).”
• Recommendations as Treatments: Debiasing Learning and Evaluation (2016, methodology): “Finally, note that other recommendation methods (e.g., Weimer et al., 2007; Lin, 2007) can in principle be adapted to propensity scoring as well.”
• Mining contextual movie similarity with matrix factorization for context-aware recommendation (2013, background): “…for ranking-oriented CF approaches that focus on the quality of recommendation lists, for instance, EigenRank [Liu and Yang 2008], CoFiRank [Weimer et al. 2007], probabilistic latent preference analysis [Liu et al. 2009], Bayesian personalized ranking [Rendle et al. 2009a], and…”
• Neural Attentive Session-based Recommendation (2017, unknown): “[1, 16, 39, 42] cannot perform well.”
• Training and testing of recommender systems on data missing not at random (2010, methodology): “There is, however, also some work on optimizing the ranking of items, e.g., measured in terms of normalized Dis-counted Cumulative Gain (nDCG) [18].”
• CLiMF: learning to maximize reciprocal rank with collaborative less-is-more filtering (2012, methodology): “The approaches proposed in this branch of LTR approximate the optimization of IR measures either by minimizing convex upper bounds of loss functions that are based on the evaluation measures [5, 31, 32], e.”
• Discrete Collaborative Filtering (2016, methodology): “In this way, recommendation by CF naturally falls into a similarity search problem—top-K item recommendation for a user can be cast into finding the top-K similar items queried by the user [30, 3, 12].”
• Model recommendation: Generating object detectors from few samples (2015, methodology): “Many matrix factorization techniques can be considered in this context, such as restricted Boltzmann machines [47], sparse coding [32, 31], and maximum margin matrix factorization optimizing directly for ranking scores [54].”
• Trace Norm Regularization: Reformulations, Algorithms, and Multi-Task Learning (2010, methodology): “To overcome the nonsmoothness, nonconvex smooth reformulations of (1) have been proposed that are solved by conjugate gradient or alternating minimization method [40, 47, 48].”
• Multiple domain user personalization (2011, background): “To prevent information overload, many collaborative filtering and recommendation systems [1, 19, 8, 3, 2] have been developed to filter the information stream and display only information the user may be interested in.”
• Jointly modeling aspects, ratings and sentiments for movie recommendation (JMARS) (2014, background): “One may argue that this is only part of a solution, since the relative values of the entries matter more than their absolute value [25, 24].”
• Learning to rank recommendations with the k-order statistic loss (2013, unknown): “Methods like CofiRank [7], CLiMF [5], or Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citati”
• TFMAP: optimizing MAP for top-n context-aware recommendation (2012, methodology): “In the case of CF, CoFiRank [34] introduced a matrix factorization method where structured estimation was used to minimize over a convex upper bound of NDCG.”
• Label Partitioning For Sublinear Ranking (2013, methodology): “For collaborative filtering, a set of items is ranked and a variety of algorithms have been proposed for this task which typically score each item in turn, for example itembased CF (Sarwar et al., 2001), latent ranking models (Weimer et al., 2007), or SVD-based systems.”
• Temporal Relational Ranking for Stock Prediction (2018, methodology): “The similar max-margin loss has been used in several applications such as recommendation [45] and knowledge-based completion [40], and demonstrated good performance in ranking tasks.”
• Ranking with non-random missing ratings: influence of popularity and positivity on evaluation metrics (2012, background): “Recently, [13, 9] measured their ranking performance in terms on Normalized Discounted Cumulative Gain (NDCG, see [5]), ignoring”