The BellKor solution to the Netflix Prize

By Robert M. Bell, Yehuda Koren, C. Volinsky. 2007

239 citations recorded by the source corpus.

How researchers see this paper

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

In Recommender Systems

Citations in the corpus

Natural Language Processing (Almost) from Scratch (2011, background): “Constructing ensembles of classifiers is a proven way to trade computational efficiency for generalization performance (Bell et al., 2007).”
Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering (2016, positive): “Collaborative Filtering (CF), especially Matrix Factorization approaches, have seen wide success at accurately modeling users’ preferences, perhaps most notably for the Netflix Prize [2, 4, 20].”
VBPR: Visual Bayesian Personalized Ranking from Implicit Feedback (2015, background): “Despite the great success, they suffer from cold start issues due to the sparsity of real-world datasets.”
Improving Aggregate Recommendation Diversity Using Ranking-Based Techniques (2012, methodology): “More details on variations of matrix factorization techniques used in recommender systems can be found in [4], [5], [30], [52], [55].”
A hierarchical model for ordinal matrix factorization (2011, methodology): “A minimal model was used to keep the message as clean as possible, although the results could plausibly be improved by blending with other models, as was shown by (Bell et al., 2007), and by taking temporal effects into account (Koren, 2009, Töscher et al.”
A Distributed, Architecture-Centric Approach to Computing Accurate Recommendations from Very Large and Sparse Datasets (2011, background): “Suggestions based on a consensus of algorithms are often more accurate than any individual approach [12,14,56,107], but combining different recommendations is usually a post-”
Fusing Similarity Models with Markov Chains for Sparse Sequential Recommendation (2016, methodology): “There have been a variety of such algorithms including Bayesian methods [9, 10], Restricted Boltzmann Machines [11], Matrix Factorization (MF) methods (the basis of many state-of-the-art recommendation approaches such as [12], [13], [14]), and so on.”
Contextual Bandits with Latent Confounders: An NMF Approach (2016, background): “This is further corroborated by the fact that the Netflix data-set, which has more than 100 million movie ratings, can be approximated surprisingly well by matrices of rank as low as 20 [7].”
From Dependence to Causation (2016, methodology): “Ensembling is a powerful technique: the winning entry of the $1,000,000 Netflix Prize was a combination of more than 100 different weak learners (Bell et al., 2008).”
The BellKor Solution to the Netflix Grand Prize (2009, methodology): “To put this in some perspective, this result is even better than those reported [1, 2, 11, 15] by using hybrid approaches such as applying a neighborhood approach on residuals of other algorithms.”
Dimensionality Reduction via Matrix Factorization for Predictive Modeling from Large, Sparse Behavioral Data (2015, methodology): “This technique was famously used to win the Netflix prize (Bell et al. 2007).”
Generating Text with Recurrent Neural Networks (2011, methodology): “The numerous mixture components of PAQ were chosen because they improved performance on a development set, so in this respect PAQ is similar in model complexity to the winning entry of the netflix prize (Bell et al., 2007).”