Numerical Comparison of Line Search Criteria in Nonlinear Conjugate Gradient Algorithms

Adeleke, O. J. and Aderemi, O. A. and Omoregbe, N. A. and Adekunle, R. A. (2013) Numerical Comparison of Line Search Criteria in Nonlinear Conjugate Gradient Algorithms. International Journal of Mathematics and Statistics Studies, 2 (1). pp. 14-24.

PDF Download (414kB)

Abstract

One of the open problems known to researchers on the application of nonlinear conjugate gradient methods for addressing unconstrained optimization problems is the influence of accuracy of linear search procedure on the performance of the conjugate gradient algorithm. Key to any CG algorithm is the computation of an optimalstep size for which many procedures have been postulated. In this paper, we assess and compare the performance of a modified Armijo and Wolfe line search procedures on three variants of nonlinear CGM by carrying out a numerical test. Experiments reveal that our modified procedure and the strong Wolfe procedures guaranteed fast convergence.

Item Type:	Article
Uncontrolled Keywords:	Nonlinear Conjugate Gradient Method, Unconstrained Optimization Problems, Armijo Line Search, Wolfe Line Search, Large Scale Problems
Subjects:	H Social Sciences > HA Statistics Q Science > QA Mathematics Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science
Depositing User:	Mr Solomon Bayoko
Date Deposited:	07 Nov 2014 10:38
Last Modified:	07 Nov 2014 10:38
URI:	http://eprints.covenantuniversity.edu.ng/id/eprint/2927

Actions (login required)

View Item