WSEAS Transactions on Mathematics
Print ISSN: 1109-2769, E-ISSN: 2224-2880
Volume 18, 2019
A New Formulation for Quasi-Newton Methods
Author:
Abstract: We develop a framework for the construction of multi-step quasi-Newton methods which utilize values of the objective function. The model developed here is constructed via interpolants of the m+1 most recent iterates / gradient evaluations, and possesses double free parameters which introduce an additional degree of flexibility. This permits the interpolating polynomials to exploit function-values which are readily available at each iteration. A new algorithm is derived for which function values are incorporated in the update of the inverse Hessian approximation at each iteration, in an attempt to accelerate convergence. The idea of incorporating function values is not new within the context of quasi-Newton methods but the presentation made in this paper presents a new approach for such algorithms. It has been shown in several earlier works that Including function values data in the update of the Hessian approximation numerically improves the convergence of Secant-like methods. The numerical performance of the new method is assessed with promising results.
Search Articles
Keywords: Unconstrained optimization, quasi-Newton methods, multi-step methods, function value algorithms, nonlinear programming, Newton method
Pages: 79-84
WSEAS Transactions on Mathematics, ISSN / E-ISSN: 1109-2769 / 2224-2880, Volume 18, 2019, Art. #10