A globally convergent BFGS method with nonmonotone line search for non-convex minimization (Q1026434)

A modified nonmonotone BFGS (Broyden-Fletcher-Goldfarb-Shanno) method is developed for solving unconstrained optimization problems. In the proposed method, the function value at each iteration allows for an occasional increase. The method can converge to a local optimal point without a convex assumption on the objective functions. The global convergence is established under certain conditions. Some numerical experimental results are presented to compare its performance with the monotone BFGS method.