Abstract:

In this paper, we develop a nonmonotonic trust region algorithm for unconstrained optimization. Different from the traditional nonmonotonic trust region algorithms, this algorithm includes a memory model, which makes the algorithm more farsighted in the sense that its behavior is not completely domina  ted by the local nature of the objective function. We present a nonmonotonic trust region algorithm that has this feature, and prove its global convergenc under suitable conditions.

Key words: unconstrained optimization, memory model, nonmonotonic trust region algorithm, global convergence