A SAEM algorithm for matrix completion problems
Keywords:
Matrix completion, EM algorithm, SAEM algorithm, collaborative filtering, principal components analysisAbstract
In this work we dealt with matrix completion problem. This problem arises in different fields, for example, systems and control theory, image processing and collaborative filtering. Given a probabilistic matrix factorization model, we present an approach based on Bayesian statistics and a stochastic expectation maximization algorithm to retrieve an array of data from a sample of its inputs. The proposed method does not require regularization parameters and estimates the rank of the matrix, in contrast to the BPMF method. The results show that the proposed method outperforms the rank of the matrix comparing to an augmented lagrangian algorithm and it is more efficient than the BPMF method.
Downloads
References
Cand`es, E., y Recht, B. (2009). Exact matrix completion via convex optimization. Foundations of Computational Mathematics, 9(6), 717-772.
Chen, M., Lin, Z., y Ma, Y. (2010). The augmented lagrange multiplier method for exact recovery of corrupted low-rank matrices. arXiv preprint arXiv:1009.5055.
Delyon, B., Lavielle, M., y Moulines, E. (1999). Convergence of a stochastic approximation version of the em algorithm. The Annals of Statistics, 27(1), 94-128.
Dempster, A., Laird, N., y Rubin, D. (1977). Maximum likelihood from incomplete data via the em algorithm. Journal of the Royal Statistical Society. Series B (Methodological)., 39(1), 1- 38.
Hassan, R. (2009). Problemas de completaci´on de matrices parcia- les (Tesis Doctoral no publicada). Universidad Polit´ecnica de Valencia.
Kuhn, E., y Lavielle, M. (2004). Coupling a stochastic approximation version of em with an mcmc procedure. ESAIM: Probability and Statistics, 9, 115-131.
Lara, H., Oviedo, H., y Yuan, J. (2014). Matrix completion via a low rank factorization model and an augmented lagrangean succesive overrelaxation algorithm. CompAMa, 2(2), 21-46.
Lawrence, N., y Urtasun, R. (2009). Non-linear matrix factorization with gaussian processes. En Proceedings of the 26th annual international conference on machine learning, pp. 601–608.
L´eger, F., Yu, G., y Sapiro, G. (2010). Efficient matrix completion with gaussian models. arXiv preprint arXiv:1010.4050.
Salakhutdinov, R., y Mnih, A. (2008a). Bayesian probabilistic matrix factorization using markov chain monte carlo. Proceedings of the International Conference in Machine Learning. ACM, 880- 887.
Salakhutdinov, R., y Mnih, A. (2008b). Probabiltic matrix factorization. Advances in neural information processing systems, 1257-1264.
Wen, Z., Yin, W., y Zhang, Y. (2012). Solving a low-rank factorization model for matrix completion by a nonlinear successive over-relaxation algorithm. Mathematical Programming Compu- tation, 4(4), 333–361.
Downloads
Published
Issue
Section
License
The opinions expressed by the authors do not necessarily reflect the position of the publisher of the publication or of UCLA. The total or partial reproduction of the texts published here is authorized, as long as the complete source and the electronic address of this journal are cited.
The authors fully retain the rights to their works, giving the journal the right to be the first publication where the article is presented. The authors have the right to use their articles for any purpose as long as it is done for non-profit. Authors are recommended to disseminate their articles in the final version, after publication in this journal, in the electronic media of the institutions to which they are affiliated or personal digital media.
