Abstract
We introduce a new method for solving Nonlinear Least Squares problems when the Jacobian matrix of the system is large and sparse.
The main features of the new method are the following:
-
a)
The Gauss-Newton equation is “partially” solved at each iteration using a preconditioned Conjugate Gradient algorithm.
-
b)
The new point is obtained using a two-dimensional trust region scheme, similar to the one introduced by Bulteau and Vial.
We prove global and local convergence results and we present some numerical experiments.
Zusammenfassung
Eine neue Methode zur Lösung nichtlinearer Least-Squares-Probleme bei hochdimensionaler dünnbesetzter Jakobimatrix wird vorgestellt.
Die wichtigsten Charakteristika sind:
-
a)
Die Gauß-Newton-Gleichung wird „teilweise” bei jeder Iteration gelöst, wobei eine präkonditionierte konjugierte Gradientenmethode verwendet wird.
-
b)
Die neue Lösung wird über eine zweidimensionale trust-region Technik erreicht, ähnlich der von Bulteau und Vial vorgeschlagenen Variante.
Globale und lokale Konvergenzaussagen werden bewiesen und anhand einiger numerischer Beispiele demonstriert.
Similar content being viewed by others
References
Aziz, M. A., Jennings, A.: A robust incomplete Cholesky-conjugate gradient algorithm. Internat. J. Numer. Methods Engin.20, 949–966 (1984).
Axelsson, O.: Incomplete block matrix factorizations preconditioning methods: The ultimate answer? J. Comput. Appl. Math.12/13, 3–18 (1985).
Axelsson, O., Brinkkemper, S., Il'in, V. P.: On some versions of incomplete block-matrix factorizations iterative methods. Linear Alg. Applics.58, 3–15 (1984).
Axelsson, O., Lindskog, G.: On the eigenvalue distribution of a class of preconditioning methods. Numer. Math.48, 479–498 (1986).
Axelsson, O., Lindskog, G.: On the rate of convergence of the preconditioned conjugate gradient method. Numer. Math.48, 499–524 (1986).
Bishop, T. N. et al.: Tomographic determination of velocity and depth in laterally varying media. Geophysics50, 903–923 (1985).
Boggs, P. T.: The convergence of the Ben Israel iteration for nonlinear least squares problems. Math. Comput.30, 512–522 (1976).
Bulteau, J. P., Vial, J. P.: A restricted trust region algorithm for unconstrained optimization. JOTA47, 413–435 (1985).
Coleman, T. F.: Large sparse numerical optimization. Lectures Notes in Computer Science No. 165. Springer-Verlag, 1984.
Dembo, R. S., Eisenstat, S. C., Steihaug, T.: Inexact Newton methods. SIAM J. of Numer. Anal.19, 400–408 (1982).
Dembo, R. S., Steihaug, T.: Truncated Newton methods. Math. Programming26, 190–212 (1983).
Dennis, J. E.: Some computational techniques for the nonlinear least-squares problems. In: Numerical Solution of Systems of Nonlinear Algebraic Equations (Byrne, G. D., Hall, C. A., eds.), pp. 157–183. New York: Academic Press 1973.
Dennis, J. E., Gay, D. M., Welsch, R. E.: An Adaptive Nonlinear Least-Squares Algorithm. Report TR 77-321. Department of Computer Sciences, Cornell University, 1977.
Dennis, J. E., Marwil, E. S.: Direct secant updates of matrix factorizations. Math. Comput.38, 459–476 (1982).
Deuflhard, P., Apostolescu, V.: A study of the Gauss-Newton method for the solution of nonlinear least squares problems. In: Special Topics of Applied Mathematics (Frehse/Pallaschke/Trottenberg, eds.), pp. 129–150. North-Holland Publ. Co., 1980.
Dennis, J., Schnabel, R.: Numerical methods for unconstrained optimization and nonlinear equations. Prentice-Hall Series in Comp. Math. N. J.: Prentice Hall 1983.
Dongarra, J. J., Leaf, G. K., Minkoff, M.: A Preconditioned Conjugate Gradient Method for Solving a Class of Non-Symmetric Linear Systems ANL-81-71. Argonne, Il. Argonne National Laboratory 1981.
Fletcher, R.: A Modified Marquardt Subroutine for Nonlinear Least Squares. Harwell: AERE Report 6799, 1971.
George, A., Heath, M. T.: Solution of sparse linear least squares using Givens rotations. Linear Algebra and Applics.34, 69–83 (1980).
George, A., Heath, M. T., Ng, E.: A comparison of some methods for solving sparse least-squares problems. SIAM J. Sci. St. Comp.4, 177–187 (1983).
Gill, P., Murray, W., Wright, M. H.: Practical Optimization. N. Y.: Academic Press 1981.
Gill, P., Murray, W., Saunders, M., Wright, M.: Sparse methods in optimization. SIAM J. Sci. St. Comp.5, 562–589 (1984).
Harwell Subroutine Library: Harwell: Subroutine NS03A, 1971.
Heath, M. T.: Numerical methods for large sparse linear least squares problems. SIAM J. Sci. St. Comp.5, 497–513 (1984).
Hestenes, M. R., Stiefel, E.: Methods of conjugate gradients for solving linear systems. J. Res. Nat. Bur. Standards49, 409–436 (1952).
Hiebert, K. L.: An evaluation of mathematical software that solves nonlinear least squares problems. ACM TOMS7, 1–16 (1981).
Le, D.: A fast and robust unconstrained optimization method requiring minimum storage. Math. Prog.32, 41–48 (1985).
Leite, L. W. B., Leão, J. W. D.: Ridge regression applied to the inversion of two-dimensional aeromagnetic anomalies. Geophysics50, 1294–1306 (1985).
Levenberg, K.: A method for the solution of certain nonlinear problems in least squares. Quarterly J. of Appl. Math..2, 164–168 (1944).
Luenberger, D. G.: Linear and Nonlinear Programming (2nd edition). Addison Wesley 1984.
Manteuffel, T. A.: Shifted incomplete factorization. In: Sparse Matrix Proceedings. (Duff, I., Stewart, G., eds.), Philadelphia: SIAM 1979.
Marquardt, D. W.: An algorithm for least-squares estimation of nonlinear parameters. SIAM J. Appl. Math.11, 431–441 (1963).
Martínez, J. M.: A quasi-Newton method with a new updating for the LDU factorization of the approximate Jacobian. Mat. Aplic. e Comput.2, 131–142 (1983).
Martínez, J. M.: A quasi-Newton method with modification of one column per iteration. Computing33, 353–362 (1984).
Martínez, J. M.: Solving systems of nonlinear equations by means of an accelerated successive orthogonal projections method. J. Comput. Appl. Math.16, 169–179 (1986).
Martínez, J. M., Sampaio, R. J. B.: Parallel and sequential Kaczmarz methods for solving underdetermined nonlinear equations. J. Comput. Appl. Math.15, 311–321 (1986).
Menzel, R.: On solving nonlinear least-squares problems in case of Rank deficient Jacobians. Computing34, 63–72 (1985).
Meyn, K. H.: Solution of underdetermined nonlinear equations by stationary iteration methods. Numer. Math.42, 161–172 (1983).
Moré, J. J.: The Levenberg-Marquardt algorithm: Implementation and theory. Proc. Biennial Conf. Num. An., Dundee 1977 (Watson, G. A., ed.). Lecture Notes in Math. Springer-Verlag (1978).
Moré, J. J.: Recent developments in algorithms and software for trust region methods. TR ANL/MCS-TM-2. Argonne: Argonne National Laboratory 1982.
Moré, J. J., Garbow, B. S., Hillstrom, K. E.: Testing unconstrained minimization software. ACM TOMS7, 17–41 (1981).
Nazareth, L.: Some recent approaches to solving large residual nonlinear least-squares problems. SIAM Review22, 1–11 (1980).
Ortega, J. M., Rheinboldt, W. C.: Iterative Solution of Nonlinear Equations in Several Variables. New York: Academic Press 1970.
Paige, C. C., Saunders, M. A., LSQR: An algorithm for sparse linear equations and sparse least squares. ACM TOMS8, 43–71 (1982).
Pedersen, L. B.: Interpretation of potential field data. A generalized inverse approach. Geophys. Prosp.25, 199–230 (1977).
Saunders, M. A.: Sparse least squares by conjugate gradients: A comparison of preconditioning methods. Proc. Computer Science and Statistics: 12th Annual Symposium on the Interface (Gentleman, J. F., ed.), pp. 15–20. Ontario: Univ. of Waterloo 1979.
Schabak, R.: Convergence analysis of the general Gauss-Newton algorithm. Numer. Math.46, 281–309 (1985).
Schubert, L. K.: Modification of a quasi-Newton method for nonlinear equations with a sparse Jacobian. Math. Comput.24, 27–30 (1970).
Schwetlick, H.: Numerische Lösung nichtlinearer Gleichungen. Berlin: Deutscher Verlag der Wissenschaften 1978.
Shanno, D. F.: Globally convergent conjugate gradient algorithms. Math. Prog.33, 61–67 (1985).
Steihaug, T.: The conjugate gradient method and trust regions in large scale optimization. SIAM J. Num. Anal.20, 626–637 (1983).
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Martínez, J.M. An algorithm for solving sparse Nonlinear Least Squares problems. Computing 39, 307–325 (1987). https://doi.org/10.1007/BF02239974
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02239974