Fast maximum margin matrix factorization for collaborative prediction

By Jason D. M. Rennie, Nati Srebro. 2005

1,119 citations recorded by the source corpus.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Recommender Systems, Statistical Learning Theory

How researchers see this paper

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

In Computational Mechanics

In Computer Vision and Pattern Recognition

In Recommender Systems

In Statistical Learning Theory

Citations in the corpus

A Neural Autoregressive Approach to Collaborative Filtering (2016, methodology): “There are other MF-based CF methods such as (Rennie & Srebro, 2005; Mackey et al., 2011).”
Explainable Recommendation: A Survey and New Perspectives (2018, methodology): “In this section we introduce explainable recommendation with user-based and item-based collaborative filtering (Resnick et al. [1994], Sarwar et al.”
Spectral Regularization Algorithms for Learning Large Incomplete Matrices (2010, background): “Netflix’s own algorithm has an RMSE of 0.9525 and the contest goal is to improve this by 10%, or an RMSE of 0.8572.”
BPR: Bayesian Personalized Ranking from Implicit Feedback (2009, methodology): “We have chosen the two diverse model classes of matrix factorization [5, 12] and learned k-nearest-neighbor [8].”
Exact Matrix Completion via Convex Optimization (2008, background): “In the area of recommender systems, users submit ratings on a subset of entries in a database, and the vendor provides recommendations based on the user’s preferences [30, 34].”
Introduction to Online Convex Optimization (2016, background): “The matrix completion model has been extremely popular since its inception in the context of recommendation systems (103; 88; 94; 70; 26; 101).”
SoRec: social recommendation using probabilistic matrix factorization (2008, methodology): “4.3 Comparison In this section, in order to show the performance improvement of our Social Recommendation (SoRec) algorithm, we compare our algorithm with some state-of-the-art algorithms: Maximum Margin Matrix Factorization (MMMF) [15], Probabilistic Matrix Factorization (PMF) [18], and Constr”
Recommender Systems (2016, positive): “The early works [133, 252, 300, 500, 569, 666] showed how different forms of matrix factorization could be used for recommendations.”
Low-rank Solutions of Linear Matrix Equations via Procrustes Flow (2015, background): “Low rank models are ubiquitous in machine learning, and over a decade of research has been dedicated to determining when such models can be eﬃciently recovered from partial information [Faz02, RS05, CR09].”
Functional matrix factorizations for cold-start recommendation (2011, methodology): “In [22], a convex relaxation for low rank matrix factorization is proposed and applied to collaborative ﬁlter-ing.”
Pairwise interaction tensor factorization for personalized tag recommendation (2010, methodology): “Also for the related task of item prediction, factorization models are known [20, 6, 16, 17] to outperform models like k-nearest-neighbour collaborative filtering or the Bayesian models URP [15] and PLSA [4].”
Recommender systems with social regularization (2011, methodology): “Recently, due to the eﬃciency in dealing with large datasets, several low-dimensional matrix approximation methods [28, 30, 31, 34] have been proposed for collaborative ﬁltering.”
Canonical Tensor Decomposition for Knowledge Base Completion (2018, methodology): “The trace norm (or nuclear norm) has been proposed as a convex relaxation of the rank (Srebro et al., 2005) for matrix completion in the setting of rating prediction, with strong theoretical guarantees (Candès & Recht, 2009).”
Understanding the Sparsity: Augmented Matrix Factorization with Sampled Constraints on Unobservables (2014, methodology): “Three popular and state-ofthe-art MF algorithms were chosen for performance comparisons, which are NMF in [14], PMF in [18] and fast MMMF in [17].”
Pedestrian Recognition with a Learned Metric (2010, background): “Nevertheless it is always possible to replace the standard hinge loss with a smooth approximation [10].”
Inferring Latent Network from Cascade Data for Dynamic Social Recommendation (2016, methodology): “An efficient and effective approach to recommender systems is to factorize the user-item rating matrix, and utilize the factorized user-specific and item-specific matrices to make further missing data prediction [20] [22] [24] [25] [34].”