A spectral PRP conjugate gradient methods for nonconvex optimization problem based on modified line search. (English) Zbl 1229.90209

Summary: A new spectral PRP conjugate gradient algorithm is developed for solving nonconvex unconstrained optimization problems. The search direction in this algorithm is proved to be a sufficient descent direction of the objective function independent of line search. To rule out possible unacceptably short step in the Armijo line search, a modified Armijo line search strategy is presented. The obtained step length is improved by employing the properties of the approximate Wolfe conditions. Under some suitable assumptions, the global convergence of the developed algorithm is established. Numerical experiments demonstrate that this algorithm is promising.


90C30 Nonlinear programming
90C53 Methods of quasi-Newton type
65K05 Numerical mathematical programming methods
Full Text: DOI