Abstract
This paper presents two quadratic regularization methods with finite-difference gradient approximations for smooth unconstrained optimization problems. One method is based on forward finite-difference gradients, while the other is based on central finite-difference gradients. In both methods, the accuracy of the gradient approximations and the regularization parameter in the quadratic models are jointly adjusted using a nonmonotone acceptance condition for the trial points. When the objective function is bounded from below and has Lipschitz continuous gradient, it is shown that the method based on forward finite-difference gradients needs at most \({\mathcal{O}}\left( n\epsilon ^{-2}\right) \) function evaluations to generate a \(\epsilon \)-approximate stationary point, where n is the problem dimension. Under the additional assumption that the Hessian of the objective is Lipschitz continuous, an evaluation complexity bound of the same order is proved for the method based on central finite-difference gradients. Numerical results are also presented. They confirm the theoretical findings and illustrate the relative efficiency of the proposed methods.
Similar content being viewed by others
Data Availability
The data defining the test problems considered in this work are available in [16].
Notes
The MATLAB/Octave codes of the test problems are freely available in the websites https://www.mat.univie.ac.at/~neum/glopt/test.html#test_unconstr and https://people.sc.fsu.edu/~jburkardt/octave_src/test_nonlin/test_nonlin.html.
The derivative-free trust-region method in [5] is designed to minimize functions of the form \(f(x)=h(F(x))\), where \(F:{\mathbb {R}}^{n}\rightarrow {\mathbb {R}}^{m}\) and \(h:{\mathbb {R}}^{m}\rightarrow {\mathbb {R}}\) is convex. In the code DFNLS we consider \(h(z)=\Vert z\Vert _{2}^{2}\). The parameters are the same considered in [5].
The data profiles were generated using the code data_profile.m freely available in the website https://www.mcs.anl.gov/~more/dfo/.
References
Larson, J., Menickelly, M., Wild, S.M.: Derivative-free optimization. Acta Numer., pp 287–404 (2019)
Vicente, L.N.: Worst case complexity of direct search. EURO J. Comput. Optim. 1, 143–153 (2013)
Konecny, J., Richtárik, P.: Simple complexity analysis of simplified direct search. arXiv:1410.0390 [math.OC] (2014)
Dodangeh, M., Vicente, L.N., Zhang, Z.: On the optimal order of worst case complexity of direct search. Optim. Lett. 10, 699–708 (2016)
Grapiglia, G.N., Yuan, J., Yuan, Y.: A derivative-free trust-region algorithm for composite nonsmooth optimization. Comput. Appl. Math. 35, 475–499 (2016)
Garmanjani, R., Júdice, D., Vicente, L.N.: Trust-region methods without using derivatives: worst-case complexity and the non-smooth case. SIAM J. Optim. 26, 1987–2011 (2016)
Nesterov, Y., Spokoiny, V.: Random gradient-free minimization of convex functions. Found. Comput. Math. 17, 527–566 (2017)
Bergou, E.H., Gorbunov, E., Richtárik, P.: Stochastic three points method for unconstrained smooth minimization. SIAM J. Optim. 30, 2726–2749 (2020)
Gratton, S., Royer, C.W., Vicente, L.N., Zhang, Z.: Direct search based on probabilistic descent. SIAM J. Optim. 25, 1515–1541 (2015)
Kimiaei, M., Neumaier, A.: Efficient global unconstrained black box optimization. Optimization Online (2021)
Cartis, C., Roberts, L.: Scalable subspace methods for derivative-free nonlinear least-squares optimization. arXiv:2102.12016 [math.OC] (2021)
Nocedal, J., Wright, S.J.: Numerical Optimization, 2nd edn. Springer, Berlin (2006)
Grapiglia, G.N., Gonçalves, M.L.N., Silva, G.N.: A cubic regularization of Newton’s method with finite-difference Hessian approximations. Numer. Algorithms 90, 607–630 (2022)
Cartis, C., Gould, N.I.M., Toint, Ph.L.: On the oracle complexity of first-order and derivative-free algorithms for smooth nonconvex minimization. SIAM J. Optim. 22, 66–86 (2012)
Cartis, C., Gould, N.I.M., Toint, Ph.L.: Adaptive cubic regularisation methods for unconstrained optimization: Part I: motivation, convergence and numerical results. Math. Progr. 127, 245–295 (2011)
Moré, J.J., Garbow, B.S., Hillstrom, K.E.: Testing unconstrained optimization software. ACM Trans. Math. Softw. 7, 17–41 (1981)
Nelder, J.A., Mead, R.: A simplex method for function minimization. Comput. J. 7, 308–313 (1965)
Grapiglia, G.N., Sachs, E.W.: On the worst-case evaluation complexity of non-monotone line search algorithms. Comput. Optim. Appl. 68, 555–577 (2017)
Moré, J.J., Wild, S.M.: Benchmarking derivative-free optimization algorithms. SIAM J. Optim. 20, 172–191 (2009)
Berahas, A.S., Byrd, R.H., Nocedal, J.: Derivative-free optimization of noisy functions via quasi-Newton methods. SIAM J. Optim. 29, 965–993 (2019)
Berahas, A.S., Cao, L., Choromanski, K., Scheinberg, K.: A theoretical and empirical comparison of gradient approximations in derivative-free optimization. Found. Comput. Math. 22, 507–560 (2021)
Berahas, A.S., Sohab, O., Vicente, L.N.: Full-low evaluation methods for derivative-free optimization. arXiv:2107.11908 [math.OC] (2021)
Shi, H-J.M., Xuan, M.Q., Oztoprak, F., Nocedal, J.: On the numerical performance of derivative-free optimization methods based on finite-difference approximations. arXiv:2102.09762 [math.OC] (2021)
Acknowledgements
The author is very grateful to two anonymous referees, whose comments helped to improve the paper.
Author information
Authors and Affiliations
Corresponding author
Additional information
This work is dedicated to Stela Angelozi Leite.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
G. N. Grapiglia was partially supported by CNPq - Brazil Grant 312777/2020-5.
Rights and permissions
About this article
Cite this article
Grapiglia, G.N. Quadratic regularization methods with finite-difference gradient approximations. Comput Optim Appl 85, 683–703 (2023). https://doi.org/10.1007/s10589-022-00373-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10589-022-00373-z