Abstract
Two new methods for unconstrained optimization are presented. Both methods employ a hybrid direction strategy which is a modification of Powell's 1970 dogleg strategy. They also employ a projection technique introduced by Davidon in his 1975 algorithm which uses projection images of Δx and Δg in updating the approximate Hessian. The first method uses Davidon's optimally conditioned update formula, while the second uses only the BFGS update. Both methods performed well without Powell's special iterations and singularity safeguards, and the numerical results are very promising.
Similar content being viewed by others
References
Powell, M. J. D.,A Hybrid Method for Nonlinear Equations, Numerical Methods for Nonlinear Algebraic Equations, Edited by P. Rabinowtiz, Gordon and Breach, London, England, 1970.
Powell, M. J. D.,A Fortran Subroutine for Solving Systems of Nonlinear Algebraic Equations, Numerical Methods for Nonlinear Algebraic Equations, Edited by P. Rabinowitz, Gordon and Breach, London, England, 1970.
Powell, M. J. D.,A Fortran Subroutine for Unconstrained Minimization, Requiring First Derivatives of the Objective Function, Atomic Energy Research Establishment, Harwell, England, Report No. R-6469, 1970.
Powell, M. J. D.,A New Algorithm for Unconstrained Optimization, Nonlinear Programming, Edited by J. B. Rosen, O. L. Mangasarian, and K. Ritter, Academic Press, New York, New York, 1970.
Davidon, W. C.,New Optimization Algorithms Without Linear Searches, Mathematical Programming, Vol. 9, pp. 1–30, 1975.
Fletcher, R.,A New Approach to Variable Metric Algorithms, Computing Journal, Vol. 13, pp. 317–322, 1970.
Gill, P. E., Murray, W., andPitfield, R. A.,The Implementation of Two Revised Quasi-Newton Algorithms for Unconstrained Optimization, National Physical Laboratory, Teddington, England, Report No. NAC-11, 1972.
Dixon, L. C. W.,All the Quasi-Newton Family Generate Identical Points, Journal of Optimization Theory and Applications, Vol. 10, pp. 34–40, 1972.
Author information
Authors and Affiliations
Additional information
Communicated by H. Y. Huang
This research was supported by the National Science Foundation under Grant No. GJ-40903.
Rights and permissions
About this article
Cite this article
Dennis, J.E., Mei, H.H.W. Two new unconstrained optimization algorithms which use function and gradient values. J Optim Theory Appl 28, 453–482 (1979). https://doi.org/10.1007/BF00932218
Issue Date:
DOI: https://doi.org/10.1007/BF00932218