Abstract
We investigate the computation of a sparse solution to an underdetermined system of linear equations using the Huber loss function as a proxy for the 1-norm and a quadratic error term à la Lasso. The approach is termed “penalized Huber loss”. The results of the paper allow to calculate a sparse solution using a simple extrapolation formula under a sign constancy condition that can be removed if one works with extreme points. Conditions leading to sign constancy, as well as necessary and sufficient conditions for computation of a sparse solution by penalized Huber loss, and ties among different solutions are presented.
Similar content being viewed by others
References
Aybat NS, Iyengar G (2011) A first-order smoothed penalty method for compressed sensing. SIAM J Optim 21(1):287–313
Aybat NS, Iyengar G (2012) A first-order augmented lagrangian method for compressed sensing. SIAM J Optim 22(2):429–459
Bertsimas D, Tsiksiklis J (1997) Introduction to Linear Optimization, Belmont. Athena Scientific, Massachusetts
Bryan K, Leise T (2013) Making do with less: an introduction to compressed sensing. SIAM Rev 55:547–566
Candès EJ, Tao T (2006) Decoding by linear programming. IEEE Trans Inform Theory 52:4203–4215
Chen BT, Madsen K, Zhang Sh (2005) On the characterization of quadratic splines. J Optim Theory Appl 124(1):93–111
Cheng W, Dai Y-H (2018) Gradient based method with active set strategy for \(\ell _1\) optimization. Math Comput 87(311):1283–1305
Cheng W, Dai Y-H An active set Newton-CG method for \(\ell _1\) optimization, Applied and Computational Harmonic Analysis, to appear
Cheng W, Hu QJ, Li D (2019) A fast conjugate gradient algorithm with active set prediction for \(\ell _1\) optimization. Optim Methods Softw 34(6):1277–1305
Cheng W, Chen Z, Li D (2015) Non-monotone spectral gradient method for sparse recovery. Inv Prob Imag 9(3):815–833
Cheng W, Li D (2018) A preconditioned conjugate gradient method with active set strategy for \(\ell _1\)-regularized least squares. J Oper Res Soc China 6:571–585
Chrétien S (2010) An alternating \(\ell _1\) approach to the compressed sensing problem. IEEE Sig Proc Lett 17:181–184
Chen S, Donoho D, Saunders M (1998) Atomic decomposition by basis pursuit. SIAM J Sci Comput 20:33–61
Donoho DL (2005) Compressed sensing. IEEE Trans Inform Theory 51:1289–1306
Donoho DL (2006) For most large underdetermined systems of linear equations the minimal \(\ell _1\)-norm solution is also the sparsest solution. Comm Pure Appl Math 59:797–829
Elad M (2010) Sparse and redundant representations: from theory to applications in signal and image processing. Springer, New York
Foucart S, Rauhut H (2013) A Mathematical Introduction to Compressive Sensing. Springer, New York
Fuchs JJ (2004) On sparse representations in arbitrary redundant bases. IEEE Trans Inform Theor 50(6):1341–1344
Huber PJ (1981) Robust Statistics. Wiley, New York
Kızılkale C, Chandrasekaran S, Ç. Pınar M, Gu M (2020) Gradient based adaptive restart is linearly convergent, Technical Report, http://www.ie.bilkent.edu.tr/~mustafap/pubs/Restartnew4.pdf. Accessed July 2020
Lanza A, Morigi S, Selesnick IW, Sgallari F (2019) Sparsity-inducing non-convex, non-separable regularization for convex image processing. SIAM J Imag Sci 12(2):1099–1134
Lee JD, Sun Y, Saunders MA (2014) Proximal Newton-type methods for minimizing composite functions. SIAM J Optim 24(3):1420–1443
Li W, Swetits JJ (1998) The linear \(\ell _1\) estimator and the Huber M-estimator. SIAM J Optim 8:457–475
Madsen K, Nielsen HB (1993) A finite smoothing algorithm for linear \(\ell _1\) estimation. SIAM J Optim 3:223–235
Madsen K, Nielsen HB, Pınar MÇ (1994) New characterizations of \(\ell _1\) solutions of overdetermined systems of linear equations. Oper Res Lett 16:159–166
Madsen K, Nielsen HB, Pınar MÇ (1998) A finite continuation algorithm for bound constrained quadratic programming. SIAM J Optim 9:62–83
Mangasarian OL, Meyer RR (1979) Nonlinear perturbations of linear programs. SIAM J Control Optim 17:745–752
Nesterov Yu (2004) Introductory lectures on convex optimization: a basic course. Kluwer Academic Publishers, Norwell, MA
Nesterov Yu (2005) Smooth minimization of nonsmooth functions. Math Program 103:127–152
Pınar MÇ (2019) Necessary and sufficient conditions for noiseless sparse recovery via convex quadratic splines. SIAM J Matrix Anal Appl 40(1):194–209
Selesnick I (2017) Sparse regularization via convex analysis. IEEE Trans Sig Process 65(17):4481–4494
Singaraju D, Tron R, Elhamifar E, Yang AY, Sastry SS (2012) On the lagrangian biduality of sparsity minimization, ICASSP 2012, IEEE Proceedings, 3429–3432
Wang S, Chen X, Dai W, Selesnick IW, Cai G (2018) Vector minimax concave penalty for sparse representation. Dig Sig Process 83:165–179
Wang J, Zhang F, Huang J, Wang W, Yuan C (2019) A non-convex penalty function with integral convolution approximation for compressed sensing. Sig Process 158:116–128
Wen Z, Yin W, Goldfarb D, Zhang Y (2010) A fast algorithm for sparse reconstruction based on shrinkage, subspace optimization, and continuation. SIAM J Sci Comp 32(4):1832–1857
Williams AC (1970) Complementarity theorems for linear programming. SIAM Rev 12:135–137
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
C. Kızılkale: The research of this author was supported in part by the Applied Mathematics program of the DOE Office of Advanced Scientific Computing Research under Contract No. DE-AC02-05CH11231.
Rights and permissions
About this article
Cite this article
Kızılkale, C., Pınar, M. Sparse solutions to an underdetermined system of linear equations via penalized Huber loss. Optim Eng 22, 1521–1537 (2021). https://doi.org/10.1007/s11081-020-09577-w
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11081-020-09577-w