Abstract
For minimizing a sum of finitely many proper, convex and lower semicontinuous functions over a nonempty closed convex set in an Euclidean space we propose a stochastic incremental mirror descent algorithm constructed by means of the Nesterov smoothing. Further, we modify the algorithm in order to minimize over a nonempty closed convex set in an Euclidean space a sum of finitely many proper, convex and lower semicontinuous functions composed with linear operators. Next, a stochastic incremental mirror descent Bregman-proximal scheme with Nesterov smoothing is proposed in order to minimize over a nonempty closed convex set in an Euclidean space a sum of finitely many proper, convex and lower semicontinuous functions and a prox-friendly proper, convex and lower semicontinuous function. Different to the previous contributions from the literature on mirror descent methods for minimizing sums of functions, we do not require these to be (Lipschitz) continuous or differentiable. Applications in Logistics, Tomography and Machine Learning modelled as optimization problems illustrate the theoretical achievements
Similar content being viewed by others
Data Availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
References
Ahookhosh, M.: Optimal subgradient methods: computational properties for large-scale linear inverse problems. Optim Eng 19, 815–844 (2018)
Z. Allen-Zhu, L. Orecchia: Linear coupling: An ultimate unification of gradient and mirror descent, In: C.H. Papadimitrou (ed.), Innovations in Theoretical Computer Science (ITCS 2017), Leibniz Int Pr Infor 8, Art. No. 3, 3:1–3:22 (2017)
M. Amini, F. Yousefian: An iterative regularized mirror descent method for ill-posed nondifferetiable stochastic optimization, http://arxiv.org/abs/1901.09506 (2019)
N. Azizan, B. Hassibi: A characterization of stochastic mirror descent algorithms and their convergence properties, In: Int Conf Acoust Spee (ICASSP-2019), 5167–5171 (2019)
Bach, F.: Duality between subgradient and conditional gradient methods. SIAM J Optim 25, 115–129 (2015)
Beck, A.: First Order Methods in Optimization. SIAM, Philadelphia (2017)
Beck, A., Teboulle, M.: Mirror descent and nonlinear projected subgradient methods for convex optimization. Oper Res Lett 31, 167–175 (2003)
Beck, A., Teboulle, M.: Smoothing and first order methods: a unified framework. SIAM J Optim 22, 557–580 (2012)
Ben-Tal, A., Margalit, T., Nemirovski, A.: The ordered subsets mirror descent optimization method with applications to tomography. SIAM J Optim 12, 79–108 (2001)
Boţ, R.I., Böhm, A.: An incremental mirror descent subgradient algorithm with random sweeping and proximal step. Optimization 68, 1–18 (2018)
Boţ, R.I., Böhm, A.: Variable smoothing for convex optimization problems using stochastic gradients, J Sci Comput 85:33 (2020)
J.-P. Calliess: Lipschitz optimisation for Lipschitz interpolation, In: 2017 American Control Conference (ACC2017), 17000349 (2017)
A. Defazio: A simple practical accelerated method for finite sums, In: D.D. Lee, U. von Luxburg, R. Garnett, M. Sugiyama and I.M.Guyon (eds.), Adv Neur In 29 (NIPS 2016)
Doan, T.T., Bose, S., Nguyen, D.H., Beck, C.L.: Convergence of the iterates in mirror descent methods. IEEE Contr Syst Lett 3, 114–119 (2019)
Duchi, J.C., Agarwal, A., Johansson, M., Jordan, M.I.: Ergodic mirror descent. SIAM J Optim 22, 1549–1578 (2012)
R.M. Freund, P. Grigas, R. Mazumder: AdaBoost and forward stagewise regression are first-order convex optimization methods, http://arxiv.org/abs/1307.1192 (2013)
G. Goh: Optimization with Costly Subgradients, ProQuest Dissertations Publishing, 2017.10685037 (2017)
Grad, S.-M., Wilfer, O.: A proximal method for solving nonlinear minmax location problems with perturbed minimal time functions via conjugate duality. J Glob Optim 74, 121–160 (2019)
Guigues, V.: Inexact stochastic mirror descent for two-stage nonlinear stochastic programs. Math Program 187, 533–577 (2021)
Hanzely, F., Richtárik, P.: Fastest rates for stochastic mirror descent methods. Comput Optim Appl 79, 717–766 (2021)
L.T.K. Hien, N. Gillis, P. Patrinos: Inertial block mirror descent method for non-convex non-smooth optimization, http://arxiv.org/abs/1903.01818 (2019)
Hien, L.T.K., Nguyen, C.V., Xu, H., Lu, C., Feng, J.: Accelerated randomized mirror descent algorithms for composite non-strongly convex optimization. J Optim Theory Appl 181, 541–566 (2019)
Hovhannisyan, V., Parpas, P., Zafeiriou, S.: MAGMA - multilevel accelerated gradient mirror descent algorithm for large-scale convex composite minimization. SIAM J Imaging Sci 9, 1829–1857 (2016)
Ivanova, A., Stonyakin, F., Pasechnyuk, D., Vorontsova, E., Gasnikov, A.: Adaptive mirror descent for the network utility maximization problem. IFAC-PapersOnLine 53, 7851–7856 (2020)
Juditsky, A., Kwon, J., Moulines, É.: Unifying mirror descent and dual averaging, Math Program 199, 793–830 (2023)
W. Krichene, A. M. Bayen, P. L. Bartlett: Accelerated mirror descent in continuous and discrete time, Adv Neur In 2 (NIPS 2015), 2845–2853 (2015)
G. Kunapuli, J. Shavlik: Mirror descent for metric learning: a unified approach, In: P.A. Flach, T. De Bie and N. Cristianini (eds.), Machine Learning and Knowledge Discovery in Databases - ECML PKDD 2012, Lect Notes Artif Int 7523, 859-874 (2012)
Y.-H. Li, C.A. Riofrío, V. Cevher: A general convergence result for mirror descent with Armijo line search, http://arxiv.org/abs/1805.12232 (2018)
Lu, H.: Relative continuity for non-Lipschitz nonsmooth convex optimization using stochastic (or deterministic) mirror descent. INFORMS J Optim 4, 288–303 (2019)
D.V.N. Luong, P. Parpas, D. Rueckert, B. Rustem: Solving MRF minimization by mirror descent, In: G. Bebis et al. (eds.), Advances in Visual Computing (ISVC 2012), Lect Notes Comput Sci 7431, Springer, 587–598 (2012)
S. Mahadevan, B. Liu, P. Thomas, W. Dabney, S. Giguere, N. Jacek, I. Gemp, J. Liu: Proximal reinforcement learning: a new theory of sequential decision making in primal-dual spaces, http://arxiv.org/abs/1405.6757 (2014)
H.B. McMahan: A unified view of regularized dual averaging and mirror descent with implicit updates, http://arxiv.org/abs/1009.3240v2 (2011)
P. Mertikopoulos, B. Lecouat, H. Zenati, C.-S. Foo, V. Chandrasekhar, G. Piliouras: Optimistic mirror descent in saddle-point problems - going the extra (gradient) mile, International Conference on Learning Representations (ICLR 2019), 1–23 (2019)
Mertikopoulos, P., Staudigl, M.: Stochastic mirror descent dynamics and their convergence in monotone variational inequalities. J Optim Theory Appl 179, 838–867 (2018)
K. Mishchenko: Sinkhorn algorithm as a special case of stochastic mirror descent, http://arxiv.org/abs/1909.06918 (2019)
A.V. Nazin, S. Anulova, A. Tremba: Application of the mirror descent method to minimize average losses coming by a Poisson flow, In: Proceedings of the European Control Conference (ECC14), 2194–2197 (2014)
Nedić, A., Lee, S.: On stochastic subgradient mirror-descent algorithm with weighted averaging. SIAM J Optim 24, 84–107 (2014)
Nemirovski, A.: Efficient methods for large-scale convex optimization problems. Ékon Mat Metody 2, 135–152 (1979). ((in Russian))
Nemirovski, A., Yudin, D.B.: Problem Complexity and Method Efficiency in Optimization. J. Wiley & Sons, New York (1983)
Nesterov, Y.: Primal-dual subgradient methods for convex problems. Math Program 120, 221–259 (2009)
Nesterov, Y.: Smooth minimization of non-smooth functions. Math Program 103, 127–152 (2005)
Y. Nesterov: Lectures on Convex Optimization, Springer (2018)
Nguyen, M.N., Le, T.H.A., Giles, D., Nguyen, T.A.: Smoothing techniques and difference of convex functions algorithms for image reconstruction. Optimization 69(7–8), 1601–1633 (2019)
Paulavicius, R.: J, Zilinskas: Analysis of different norms and corresponding Lipschitz constants for global optimization. Inf Technol Control 12, 301–306 (2006)
Quoc, T.-D.: Adaptive smoothing algorithms for nonsmooth composite convex minimization. Comput Optim Appl 66, 425–451 (2017)
R.T. Rockafellar: Convex Analysis, Princeton University Press (1970)
S. Roweis: Data for Matlab hackers, http://www.cs.nyu.edu/data.html
F. Schäfer, A. Anandkumar, H. Owhadi: Competitive mirror descent, http://arxiv.org/abs/2006.10179 (2020)
Semenov, V.V.: A version of the mirror descent method to solve variational inequalities. Cybern Syst Anal 53, 234–243 (2017)
Titov, A., Stonyakin, F., Alkousa, M., Ablaev, S., Gasnikov, A.: Analogues of switching subgradient schemes for relatively Lipschitz-continuous convex programming problems. MOTOR 2020, 133–149 (2020)
S. Zhang, N. He: On the convergence rate of stochastic mirror descent for nonsmooth nonconvex optimization, http://arxiv.org/abs/1806.04781 (2018)
X. Zhou, C. Du, X. Cai: An efficient smoothing proximal gradient algorithm for convex clustering, http://arxiv.org/abs/2006.12592 (2020)
Y. Zhou, Y. Liang, L. Shen: A unified approach to proximal algorithms using Bregman distance, Technical Report, Syracuse University (2016)
Zhou, Z., Mertikopoulos, P., Bambos, N., Boyd, S.P., Glynn, P.W.: On the convergence of mirror descent beyond stochastic convex programming. SIAM J Optim 30, 687–716 (2020)
J. Zimmert, T. Lattimore: Connections between mirror descent, Thompson sampling and the information ratio, In: H. Wallach, H. Larochelle, A. Beygelzimer, F. d’Alché-Buc, E. Fox and R. Garnett (eds.), Adv Neur In 32 (NIPS 2019), 11973–11982 (2019)
Acknowledgements
The work of the first named author was supported by the German Research Foundation (DFG), project WA922/9-1. The work of the second named author was partially supported by the Austrian Science Fund (FWF), project M-2045, by the Hi! PARIS Center, and by a public grant as part of the Investissement d’avenir project, reference ANR-11-LABX-0056-LMH, LabEx LMH. For useful discussions regarding this paper the authors are thankful to Radu Ioan Boţ and Axel Böhm, to whom we are also grateful for providing us the program codes for their numerical experiments from [10], and to Gert Wanka. Valuable comments and suggestions from two anonymous reviewers are gratefully acknowledged, too.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
There are no competing interests to declare.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
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
Bitterlich, S., Grad, SM. Stochastic incremental mirror descent algorithms with Nesterov smoothing. Numer Algor 95, 351–382 (2024). https://doi.org/10.1007/s11075-023-01574-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11075-023-01574-1