A PROJECTED GRADIENT FILTER TRUST-REGION ALGORITHM FOR BOUND CONSTRAINED OPTIMIZATION

作者：Fatemi M; Mahdavi Amiri N^*

来源：Pacific Journal of Optimization, 2014, 10(1): 229-242.

• bound constrained optimization
• filter methods
• trust-region algorithms
• gradient projection method

We present a filter trust-region algorithm for solving bound constrained optimization problems. The algorithm is an extension of our method recently proposed for unconstrained optimization to consider the bound constraints. The new algorithm combines the gradient projection method with the filter technique to generate non-monotone iterations. In contrast with the earlier filter algorithms, the new algorithm has the novelty to ensure the finiteness of the filter size. Global convergence to at least one first order critical point is established. Comparative numerical results on a set of test problems from the CUTEr collection show the algorithm is competitive and more efficient solely with respect to the filter size.

• 出版日期2014-1