Family weak conjugate gradient algorithms and their convergence analysis for nonconvex functions (Q2189404)

This paper presents a family of weak two-term conjugate gradient methods for nonconvex functions. They have the following properties: (1) sufficient descent properties and trust region characteristic without other assumptions; (2) global convergence of the given algorithms under normal assumptions for the nonconvex functions, and linear convergence rate and the \(n\)-step quadratic convergence rate for uniformly convex functions. Numerical results illustrate that the presented algorithms are competetive with the normal algorithms.