Smooth singular value thresholding algorithm for low-rank matrix completion problem
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The matrix completion problem is to predict missing entries of a data matrix using the low-rank approximation of the observed entries. Typical approaches to matrix completion problem often rely on thresholding the singular values of the data matrix. However, these approaches have some limitations. In particular, a discontinuity is present near the thresholding value, and the thresholding value must be manually selected. To overcome these difficulties, we propose a shrinkage and thresholding function that smoothly thresholds the singular values to obtain more accurate and robust estimation of the data matrix. Furthermore, the proposed function is differentiable so that the thresholding values can be adaptively calculated during the iterations using Stein unbiased risk estimate. The experimental results demonstrate that the proposed algorithm yields a more accurate estimation with a faster execution than other matrix completion algorithms in image inpainting problems.
Abstract
The matrix completion problem is to predict missing entries of a data matrix using the low-rank approximation of the observed entries. Typical approaches to matrix completion problem often rely on thresholding the singular values of the data matrix. However, these approaches have some limitations. In particular, a discontinuity is present near the thresholding value, and the thresholding value must be manually selected. To overcome these difficulties, we propose a shrinkage and thresholding function that smoothly thresholds the singular values to obtain more accurate and robust estimation of the data matrix. Furthermore, the proposed function is differentiable so that the thresholding values can be adaptively calculated during the iterations using Stein unbiased risk estimate. The experimental results demonstrate that the proposed algorithm yields a more accurate estimation with a faster execution than other matrix completion algorithms in image inpainting problems.
참고문헌 (21)
- H. Abdi, "Partial least squares regression and projection on latent structure regression(PLS Regression)", Comput. Stat, 2(1), pp.97~106 (2010)
- J. -F. Cai, "A singular value thresholding algorithm for matrix completion", SIAM J. Optim, 20(4), pp.1956~1982 (2010)[DOI]
- E. J. Candes, "Stable signal recovery from incomplete and inaccurate measurements", Comm. Pure Appl. Math, 59(8), pp.1207~1223 (2006)[DOI]
- E. J. Candes, "Unbiased risk estimates for singular value thresholding and spectral estimators", IEEE Trans. Signal Process, 61(19), pp.4643~4657 (2013)[DOI]
- A. Cui, "Affine matrix rank minimization problem via non-convex fraction function penalty", J. Comput. Appl. Math, 336, pp.353~374 (2018)[DOI]
- M. Fornasier, "Low-rank matrix recovery via iteratively reweighted least squares minimization", SIAM J. Optim, 21(4), pp.1614~1640 (2011)[DOI]
- M. Gavish, "Optimal shrinkage of singular values", IEEE Trans. Inform. Theory, 63(4), pp.2137~2152 (2017)[DOI]
- P. C. Hansen, "Fundamentals of Algorithms" (2010), SIAM
- P. Jain, "Guaranteed rank minimization via singular value projection", Adv. Neural Info. Proc. Syst, pp.937~945 (2010)
- J. Josse, "Adaptive shrinkage of singular values", Stat. Comput, 26(3), pp.715~724 (2016)[DOI]
- K. Lee, "ADMiRA : atomic decomposition for minimum rank approximation", IEEE Trans. Inform. Theory, 56(9), pp.4402~4416 (2010)[DOI]
- H. Y. Li, "Minimization of fraction function penalty in compressed sensing", IEEE Trans. Neural Netw. Learn. Syst, 31(5), pp.1626~1637 (2020)
- W. Li, "Non-local image inpainting using low-rank matrix completion", Comput. Graph. Forum, 34(6), pp.111~122 (2015)
- H. Liu, "Regularized partial least squares for multi-label learning", J. Mach. Learn. Cyber, 9, pp.335~346 (2018)[DOI]
- K. Mohan, "Iterative reweighted algorithms for matrix rank minimization", J. Mach. Learn. Res, 13, pp.3441~3473 (2012)
- L. T. Nguyen, "Low-rank matrix completion : a contemporary survey", IEEE Access, 7, pp.94215~94237 (2019)
- B. Recht, "Guaranteed minimum-rank solutions of linear matrix equations via nuclear norm minimization", SIAM Rev, 52(3), pp.471~501 (2010)[DOI]
- C. M. Stein, "Estimation of the mean of a multivariate normal distribution", Ann. Statist, 9(6), pp.1135~1151 (1981)
- P. Symeonidis, "Matrix and Tensor Factorization Techniques for Recommender Systems" (2016), Springer
- M. Verbanck, "Regularised PCA to denoise and visualise data", Stat. Comput, 25(2), pp.471~486 (2015)[DOI]
- Y. Wan, "Robust impulse noise variance estimation based on image histogram", IEEE Sig. Proc. Letter, 17(5), pp.485~488 (2010)