Abstract
This paper presents sum-of-squares (SOS) relaxation results to a difference-of-convex-max (DC-max) optimization involving SOS-convex polynomials in the face of constraint data uncertainty and their applications to robust feature selection. The main novelty of the present work in relation to the recent research in robust convex and DC optimization is the derivation of a new form of minimally exact SOS relaxations for robust DC-max problems. This leads to the identification of broad classes of robust DC-max problems with finitely exact SOS relaxations that are numerically tractable. They allow one to find the optimal values of these classes of DC-max problems by solving a known finite number of semi-definite programs (SDPs) for certain concrete cases of commonly used uncertainty sets in robust optimization. In particular, we derive relaxation results for a class of robust fractional programs. Also, we provide a finitely exact SDP relaxation for a DC approximation problem of an NP-hard robust feature selection model which gives computable upper bounds for the global optimal value.
Similar content being viewed by others
Data Availability
No data sets were generated or analyzed during the current study, and so data sharing is not applicable to this article.
References
Ahmadi, A.A., Parrilo, P.A.: A complete characterization of the gap between convexity and SOS-convexity. SIAM J. Optim. 23(2), 811–833 (2013)
Ben-Tal, A., El Ghaoui, L., Nemirovski, A.: Robust Optimization. Princeton University Press, Princeton (2009)
Ben-Tal, A., Nemirovski, A.: Lectures on Modern Convex Optimization: Analysis, Algorithms, and Engineering Applications. SIAM, Philadelphia (2001)
Bradley, P. S., Mangasarian, O. L.: Feature selection via concave minimization and support vector machines. In: Shavlik, J. W. (ed.) International Conference on Machine Learning (ICML), vol. 98, pp. 82–90 (1998)
Bradley, P.S., Mangasarian, O.L., Street, W.N.: A complete characterization of the gap between convexity and SOS-convexity. INFORMS J. Comput. 10(2), 209–217 (1998)
Bruckstein, A.M., Donoho, D.L., Elad, M.: From sparse solutions of systems of equations to sparse modeling of signals and images. SIAM Rev. 51(1), 34–81 (2009)
Cervantes, J., Garcia-Lamont, F., Rodrıguez-Mazahua, L., Lopez, A.: A comprehensive survey on support vector machine classification: applications, challenges and trends. Neurocomputing 408, 189–215 (2020)
Chieu, N.H., Chuong, T.D., Jeyakumar, V., Li, G.: A copositive Farkas lemma and minimally exact conic relaxations for robust quadratic optimization with binary and quadratic constraints. Oper. Res. Lett. 47(6), 530–536 (2019)
Chieu, N.H., Feng, J.W., Gao, W., Li, G., Wu, D.: SOS-convex semialgebraic programs and its applications to robust optimization: a tractable class of nonsmooth convex optimization. Set-Valued Var. Anal. 26, 305–326 (2018)
Dinkelbach, W.: On nonlinear fractional programming. Manage. Sci. 13(7), 492–498 (1967)
Dunbar, M., Murray, J.M., Cysique, L.A., Brew, B.J., Jeyakumar, V.: Simultaneous classification and feature selection via convex quadratic programming with application to HIV-associated neurocognitive disorder assessment. Eur. J. Oper. Res. 206(2), 470–478 (2010)
Gaudioso, M., Gorgone, E., Hiriart-Urruty, J.B.: Feature selection in SVM via polyhedral k-norm. Optim. Lett. 14(1), 19–36 (2020)
Gotoh, J.Y., Takeda, A., Tono, K.: DC formulations and algorithms for sparse optimization problems. Math. Program. 169, 141–176 (2018)
Grant, M., Boyd, S.: CVX: Matlab software for disciplined convex programming (2011). http://cvxr.com/cvx
Harada, R., Kuroiwa, D.: Lagrange-type duality in DC programming. J. Math. Anal. Appl. 418(1), 415–424 (2014)
Helton, J.W., Nie, J.: Semi-definite representation of convex sets. Math. Program. 120(2), 21–64 (2010)
Hiriart-Urruty, J. B., Lemarechal, C.: Fundamentals of Convex Analysis. Springer Science & Business Media (2004)
Horel, A.E.: Application of ridge analysis to regression problems. Chem. Eng. Prog. 58, 54–59 (1962)
Jeyakumar, V., Li, G.: A new class of alternative theorems for SOS-convex inequalities and robust optimization. Appl. Anal. 94(1), 56–74 (2015)
Jeyakumar, V., Lee, G.M., Linh, N.T.H.: Generalized Farkas’ lemma and gap-free duality for minimax DC optimization with polynomials and robust quadratic optimization. J. Glob. Optim. 64, 679–702 (2016)
Jeyakumar, V., Li, G.: Exact SDP relaxations for classes of nonlinear semi-definite programming problems. Oper. Res. Lett. 40(6), 529–536 (2012)
Jeyakumar, V., Li, G., Vicente-Perez, J.: Robust SOS-convex polynomial optimization problems: exact SDP relaxations. Optim. Lett. 9, 1–18 (2015)
Jeyakumar, V., Vicente-Perez, J.: Dual semi-definite programs without duality gaps for a class of convex minimax programs. J. Optim. Theory Appl. 162, 735–753 (2014)
Lasserre, J.B.: An Introduction to Polynomial and Semi-Algebraic Optimization. Cambridge University Press, Cambridge (2015)
Lasserre, J.B.: Convexity in semi-algebraic geometry and polynomial optimization. SIAM J. Optim. 19(4), 1995–2014 (2009)
Le Thi, H.A., Le, H.M., Pham Dinh, T.: Feature selection in machine learning: an exact penalty approach using a difference of convex function algorithm. Mach. Learn. 101, 163–186 (2015)
Le Thi, H. A., Pham Dinh, T.: Open issues and recent advances in DC programming and DCA. J. Glob. Optim. 1–58 (2023)
Le Thi, H.A., Vo, X.T., Pham Dinh, T.: Feature selection for linear SVMs under uncertain data: robust optimization based on difference of convex functions algorithms. Neural Netw. 59, 36–50 (2014)
Lee, J.H., Lee, G.M.: On minimizing difference of an SOS-convex polynomial and a support function over an SOS-concave matrix polynomial constraint. Math. Program. 169, 177–198 (2018)
Martınez-Legaz, J.E., Volle, M.: Duality in DC programming: the case of several DC constraints. J. Math. Anal. Appl. 237(2), 657–671 (1998)
Rockafellar, R.T.: Convex Analysis. Princeton University Press, Princeton (1970)
Su, C.T., Yang, C.H.: Feature selection for the SVM: an application to hypertension diagnosis. Expert Syst. Appl. 34(1), 754–763 (2008)
Tibshirani, R.: Regression shrinkage and selection via the Lasso. J. R. Stat. Soc. B 58(1), 267–288 (1996)
Woolnough, D., Jeyakumar, N., Li, G., Loy, C. T., Jeyakumar, V.: Robust optimization and data classification for characterization of Huntington disease onset via duality methods. J. Optim. Theory Appl. 1–27 (2022)
Zhang, W., Hong, B., Liu, W., Ye, J., Cai, D., He, X., Wang, J.: Scaling up sparse support vector machines by simultaneous feature and sample reduction. J. Mach. Learn. Res. 20(121), 1–39 (2019)
Acknowledgements
The work was partially supported by a grant from the Australian Research Council. The second author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIT) (NRF-2022R1A2C1003309). The third author was supported by the National Research Foundation of Korea (NRF) grant Funded by the Korea Government (MSIT) (NRF-2021R1C1C2004488). The authors are grateful to the referees and the handling editor for their valuable suggestions and constructive comments which have contributed to the final preparation of the paper.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The first author declares that he is an Associate Editor for the Journal of Optimization Theory and Applications. The other three authors have no conflict of interest to declare that are relevant to the content of this article.
Additional information
Communicated by Luis Zuluaga.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: Supplementary Proofs for Sect. 4
Appendix A: Supplementary Proofs for Sect. 4
Lemma A.1
Let \(\varphi _k: [0,1]^n \rightarrow {{\mathbb {R}}}\) be
Then, \(\varphi _k\) is concave.
Proof
Take \((r_1,\ldots ,r_n), (r_1',\ldots ,r_n') \in [0,1]^n\) and \(\theta \in (0,1)\). There exist \(\mu \in {{\mathbb {R}}}, {\widetilde{\xi }}_i,{\overline{\xi }}_i, \overline{p}_j, \overline{q}_j \ge 0\) such that \(\varphi (r_1,\ldots , r_n) = \mu \) and
There exist \(\mu '\in {{\mathbb {R}}}, {\widetilde{\xi }}_i',{\overline{\xi }}_i', \overline{p}_j', \overline{q}_j' \ge 0\) and
Now,
By definition, \(\varphi _k\) is concave. \(\square \)
Infimum of a concave function over the convex hull of finitely many points is equal to the infimum over the set of those points [31, Theorem 32.2].
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Jeyakumar, V., Lee, G.M., Lee, J.H. et al. Sum-of-Squares Relaxations in Robust DC Optimization and Feature Selection. J Optim Theory Appl 200, 308–343 (2024). https://doi.org/10.1007/s10957-023-02312-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10957-023-02312-2
Keywords
- DC optimization
- SOS-convex polynomials
- Robust optimization
- Semi-definite programs
- Robust feature selection