Levenberg-marquardt revisited and parameter tuning of river regression models

Springer Science and Business Media LLC - Tập 43 - Trang 1-24 - 2023
J. M. Martínez1
1Department of Applied Mathematics, Institute of Mathematics, Statistics, and Scientific Computing (IMECC), State University of Campinas, Campinas, Brazil

Tóm tắt

The Levenberg-Marquardt method is well known for solving nonlinear least squares problems. This method is mostly used in the context of overdetermined systems. In this paper it is shown that suitable implementations with respect to underdetermined systems can be defined. The resulting method is equipped with a new effective non-monotone strategy and it is proved that, when the residual tends to zero, a sufficient descent condition is obtained with minimal computational cost. The method is applied to the problem of parameter fitting of regression models based on Neural Networks for natural rivers.

Tài liệu tham khảo

Birgin EG, Correa MR, González-López V, Martínez JM, Rodrigues DS (2023) Randomly supported models for the prediction of flows in channels, Technical Report. University of Campinas Birgin EG, Martínez JM (2023) A PDE-informed optimization algorithm for river flow predictions. Numer Algorithms. https://doi.org/10.1007/s11075-023-01647-1 Bottou L, Curtis FE, Nocedal J (2018) Optimization methods for large-scale machine learning. SIAM Rev 60:223–311 Brunton SL, Kutz JL (2020) Data-driven science and engineering. Cambridge University Press, Cambridge Conn AR, Conn AR, Gould NIM, Toint PhL (2000) Trust-region methods, MPS/SIAM series on optimization. SIAM, Philadelphia Dennis JE, Schnabel RS (1983) Numerical methods for unconstrained optimization and nonlinear equations. Prentice Hall, Englewood Cliffs Emmett WW, Myrick WW, Meade RH (1979) Field data describing the movement and storage of sediment in the East Fork River, Wyoming, Part 1, River Hydraulics and Sediment Transport, Report No. 1 Golub GH, Van Loan Ch (1996) Matrix computations, 3rd edn. Johns Hopkins University Press, Baltimore Grippo L, Lampariello F, Lucidi S (1986) Nonmonotone line search technique for Newton’s method. SIAM J Numer Anal 23:707–716 Karas EW, Santos SA, Svaiter BF (2016) Algebraic rules for computing the regularization parameter of the Levenberg-Marquardt method. Comput Optim Appl 65:723–751 Levenberg K (1944) A method for the solution of certain nonlinear problems in least squares. Q Appl Math 2:164–168 LeVeque RJ (1992) Numerical methods for conservation laws. Lectures in mathematics. Birkäuser, ETH Zürich Marquardt DW (1963) An algorithm for least squares estimation of nonlinear parameters. SIAM J Appl Math 11:431–441 Martínez JM (1986) Solution of nonlinear systems of equations by an optimal projection method. Computing 37:59–70 Martínez JM (1991) Quasi-Newton methods for solving underdetermined nonlinear simultaneous equations. J Comput Appl Math 34:171–190 Meade RH, Myrick WW, Emmett WW (1979) Field data describing the movement and storage of sediment in the East Fork River, Wyoming, Part 2, River Hydraulics and Sediment Transport, Report No. 2 Meyn K-H (1983) Solution of underdetermined nonlinear equations by stationary iteration methods. Numer Math 42:161–172 Moré JJ (1978) The Levenberg-Marquardt algorithm: implementation and theory. In: Watson GA (ed) Numerical analysis. Lecture notes in mathematics, vol 630. Springer, Berlin Nocedal J, Wright SJ (1999) Numerical optimization. Springer, New York Saint-Venant AJC (1871) Théorie du mouvement non-permanent des eaux, avec application aux crues des rivière at à l’introduction des marées dans leur lit. C R Séances Acad Sci 73:147–154 Sra S, Nowozin S, Wright SJ (2011) Optimization for machine learning. The MIT Press, Cambridge Walker H, Watson LT (1990) Least-change secant update methods for under-determined systems. SIAM J Numer Anal 27:1227–1262 Wright SJ (2023) Optimization in machine learning and data science. SIAM News 56:1–2