Sample selection for MCMC-based recommender systems

Zitieren

Dateien zu dieser Ressource

Dateien Größe Format Anzeige

Zu diesem Dokument gibt es keine Dateien.

SILBERMANN, Thierry, Immanuel BAYER, Steffen RENDLE, 2013. Sample selection for MCMC-based recommender systems. the 7th ACM conference. Hong Kong, China, 12. Okt 2013 - 16. Okt 2013. In: Proceedings of the 7th ACM conference on Recommender systems - RecSys '13. the 7th ACM conference. Hong Kong, China, 12. Okt 2013 - 16. Okt 2013. New York, New York, USA:ACM Press, pp. 403-406. ISBN 978-1-4503-2409-0

@inproceedings{Silbermann2013Sampl-26477, title={Sample selection for MCMC-based recommender systems}, year={2013}, doi={10.1145/2507157.2507224}, isbn={978-1-4503-2409-0}, address={New York, New York, USA}, publisher={ACM Press}, booktitle={Proceedings of the 7th ACM conference on Recommender systems - RecSys '13}, pages={403--406}, author={Silbermann, Thierry and Bayer, Immanuel and Rendle, Steffen} }

Rendle, Steffen Bayer, Immanuel Silbermann, Thierry 2014-02-25T09:56:09Z Sample selection for MCMC-based recommender systems 2014-02-25T09:56:09Z Bayer, Immanuel deposit-license Rendle, Steffen Bayesian Inference with Markov Chain Monte Carlo (MCMC) has been shown to provide high prediction quality in recommender systems. The advantage over learning methods such as coordinate descent/alternating least-squares (ALS) or (stochastic) gradient descent (SGD) is that MCMC takes uncertainty into account and moreover MCMC can easily integrate priors to learn regularization values. For factorization models, MCMC inference can be done with efficient Gibbs samplers.<br /><br /><br />However, MCMC algorithms are not point estimators, but they generate a chain of models. The whole chain of models is used to calculate predictions. For large scale models like factorization methods with millions or billions of model parameters, saving the whole chain of models is very storage intensive and can even get infeasible in practice. In this paper, we address this problem and show how a small subset from the chain of models can approximate the predictive distribution well. We use the fact that models from the chain are correlated and propose online selection techniques to store only a small subset of the models. We perform an empirical analysis on the large scale Netflix dataset with several Bayesian factorization models, including matrix factorization and SVD++. We show that the proposed selection techniques approximate the predictions well with only a small subset of model samples. Silbermann, Thierry 2013 RecSys '13 : proceedings of the 7th ACM conference on Recommender systems ; Hong Kong, China — October 12 - 16, 2013 / Qiang Yang ... (eds.). - New York : ACM, 2013. - S. 403-406. - ISBN 978-1-4503-2409-0 eng

Das Dokument erscheint in:

KOPS Suche


Stöbern

Mein Benutzerkonto