Global Convergence of a New Coefficient Nonlinear Conjugate Gradient Method

Nur Syarafina Mohamed (Universiti Kuala Lumpur)
Mustafa Mamat (Universiti Sultan Zainal Abidin)
Mohd Rivaie (Universiti Teknologi Mara)
Shazlyn Milleana Shaharuddin (Universiti Pendidikan Sultan Idris)

Indonesian Journal of Electrical Engineering and Computer Science
Vol 11, No 3: September 2018

Nonlinear conjugate gradient (CG) methods are widely used in optimization field due to its efficiency for solving a large scale unconstrained optimization problems. Many studies and modifications have been developed in order to improve the method. The method is known to possess sufficient descend condition and its global convergence properties under strong Wolfe-Powell search direction. In this paper, the new coefficient of CG method is presented. The global convergence and sufficient descend properties of the new coefficient are established by using strong Wolfe-Powell line search direction. Results show that the new coefficient is able to globally converge under certain assumptions and theories.