Page 64 - programme book

OR-005
New Hybrid Conjugate Gradient Method under Exact Line Search

Ain Aqiela Azamuddin 1, a) , Nurul ‘Aini 1, b) , Mohd Rivaie 2, c) and Zuraida Alwadood 3, d)

1 Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA,
Cawangan Johor, Kampus Segamat, Malaysia.
2 Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA,
Cawangan Terengganu, Kampus Kuala Terengganu, Malaysia.
3 Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA,
Shah Alam, Malaysia.

a) Corresponding author: ainaqiela97@gmail.com
b) ainiharun@uitm.edu.my
c) rivaie75@uitm.edu.my
d) zuraida794@uitm.edu.my

Abstract. Conjugate gradient (CG) method is one of the popular method in solving unconstrained
optimization problem. This method is notable for being an intermediate between the steepest descent
method and the Newton’s method. In this study, a new hybrid CG method is proposed with the main
focus on improving Aini-Rivaie-Mustafa (ARM) CG method that were introduced in 2016. The ARM
CG method sometimes generates negative CG coefficient that affects the performance of the method.
Therefore, the new hybrid CG method is proposed with the intention of eliminating the negative CG
coefficient value generated by the ARM CG method. The new hybrid CG method is globally
convergent under the exact minimization rules and based on the numerical observation, it shows that
it could solve higher number of test problems compared to ARM CG method.

Keywords: Unconstrained optimization, hybrid conjugate gradient, exact line search

62