A New Hybrid Conjugate Gradient Method with Global Convergence Properties

Abstract

This work introduces a novel hybrid conjugate gradient (CG) technique for tackling unconstrained optimisation problems with improved efficiency and effectiveness. The parameter is computed as a convex combination of the standard conjugate gradient techniques using and . Our proposed method has shown that when using the strong Wolfe-line-search (SWC) under specific conditions, it achieves global theoretical convergence. In addition, the new hybrid CG approach has the ability to generate a search direction that moves downward with each iteration. The quantitative findings obtained by applying the recommended technique about 30 functions with varying dimensions clearly illustrate its effectiveness and potential. This work introduces a novel hybrid conjugate gradient (CG) technique for tackling unconstrained optimisation problems with improved efficiency and effectiveness. The parameter is computed as a convex combination of the standard conjugate gradient techniques using and . Our proposed method has shown that when using the strong Wolfe-line-search (SWC) under specific conditions, it achieves global theoretical convergence. In addition, the new hybrid CG approach has the ability to generate a search direction that moves downward with each iteration. The quantitative findings obtained by applying the recommended technique about 30 functions with varying dimensions clearly illustrate its effectiveness and potential.

Keywords

• Unconstraint Optimization
• Hybrid Conjugate Gradient
• , Line Search
• Global Convergence
• Strong Wolfe Conditions.