Revision of a derivative-free quasi-Newton method
HTML articles powered by AMS MathViewer
- by John Greenstadt PDF
- Math. Comp. 32 (1978), 201-221 Request permission
Abstract:
A derivative-free Quasi-Newton (DFQN) method previously published [J. Greenstadt, Math. Comp., v. 26, 1972, pp. 145-166] has been revised and simplified. The main modification has the effect of keeping all the successive approximants to the Hessian matrix positive-definite. This, coupled with some improvements in the line search, has enhanced the performance of the method considerably. The results of numerical trials on many of the "standard" test functions are displayed, in addition to comparisons with two other methods. These indicate that the present DFQN method is not too far behind that of Gill, Murray and Pitfield, the most efficient one presently known.References
- John Greenstadt, A quasi-Newton method with no derivatives, Math. Comp. 26 (1972), 145–166. MR 305592, DOI 10.1090/S0025-5718-1972-0305592-X
- Marvin Marcus and Henryk Minc, A survey of matrix theory and matrix inequalities, Allyn and Bacon, Inc., Boston, Mass., 1964. MR 0162808
- R. Fletcher and M. J. D. Powell, A rapidly convergent descent method for minimization, Comput. J. 6 (1963/64), 163–168. MR 152116, DOI 10.1093/comjnl/6.2.163
- H. H. Rosenbrock, An automatic method for finding the greatest or least value of a function, Comput. J. 3 (1960/61), 175–184. MR 136042, DOI 10.1093/comjnl/3.3.175 A. R. COLVILLE, A Comparative Study of Non-Linear Programming Codes, IBM NY Scientific Center Report #320-2949, 1968. M. J. D. POWELL, Comput. J., vol. 5, 1962, p. 147. R. P. BRENT, Algorithms for Finding Zeros and Extrema of Functions Without Calculating Derivatives, Stanford Univ. Comput. Sci. Report STAN-CS-71-198, 1971.
- R. Fletcher, Function minimization without evaluating derivatives—a review, Comput. J. 8 (1965), 33–41. MR 187377, DOI 10.1093/comjnl/8.1.33 M. C. BIGGS, J. Inst. Math. Appl., vol. 8, 1972, p. 315. P. E. GILL, W. MURRAY & R. A. PITFIELD, The Implementation of Two Revised Quasi-Newton Algorithms for Unconstrained Optimization, Nat. Phys. Lab. Report NAC 11, 1972. M. J. D. POWELL, Comput. J., vol. 7, 1964, p. 155.
- M. J. D. Powell, A view of unconstrained minimization algorithms that do not require derivatives, ACM Trans. Math. Software 1 (1975), 97–107. MR 411169, DOI 10.1145/355637.355638
Additional Information
- © Copyright 1978 American Mathematical Society
- Journal: Math. Comp. 32 (1978), 201-221
- MSC: Primary 65K05
- DOI: https://doi.org/10.1090/S0025-5718-1978-0474810-5
- MathSciNet review: 0474810