| Summary: | Optimization has been widely used in daily life. Optimization can be categorized into constrained and unconstrained. Unconstrained optimization can be solved using an iterative method. The iterative method can be solved by any optimization method under exact or inexact line search. Different method will lead to different search direction. One of the famous optimization methods is conjugate gradient (CG) method. LAMR+ method is one of the CG methods proposed by Linda-Aini-Mustafa-Rivaie. This method yields a good numerical performance but under some circumstances, LAMR+ method produces a higher iteration number (NOI) and CPU time. In order to overcome this problem, LAMR+ is compared under five different sets of strong Wolfe line search parameters. LAMR+ is tested using 15 test functions with different dimensions in order to identify the most effective and robust method. Four initial points are picked randomly for each test function. All the functions are tested using MatlabR2019a subroutine programming. The numerical results are compared in terms of iteration number and CPU time. Sigmaplot is used to display the performance profile of the numerical result. LAMR+ using strong Wolfe line search with parameters of 0.1 and 0.3 is the best method since it can solve almost all the test function faster than the other parameters. A regression model is formed using Terengganu Covid-19 cases for August 2021 in order to ensure the applicability of the LAMR+ method. As a conclusion, this method can be implemented for real life data. © 2024 Author(s).
|
|---|