Utilization of Novel Memoryless Conjugate Gradient Algorithms for Nonlinear Unconstrained Optimization Challenges

Abstract

This study introduces a novel family of memoryless nonlinear conjugate gradient algorithms that produce an appropriate search direction for gradient descent at each iteration. This condition is applicable irrespective of the exactness of the line search and the convexity of the goal function. We demonstrate that the offered approaches achieve global convergence for non-convex functions under specific conditions. Numerical findings illustrate the efficacy of these novel hybrid approaches when applied to specific test issues in comparison to the conventional Mawlana CG algorithm.