Comparisons of two Non-Monotone Strategies for Solving Derivative-free Wedge Trust-Region Problems

Yan Li¹, Qinghua Zhou¹, Yan Zhan²

1School of Applied Mathematics, Beijing Normal University, Zhuhai, 519087, China

2Key Lab. of Machine Learning and Computational Intelligence, College of Mathematics and Information Science, Hebei University, Baoding, 071002, China

Abstract: In this paper, we compare the differences of non-monotone strategies to solve the wedge trust re-gion method for derivative-free optimization. The non-monotone method is effective to resolve the trust re-gion algorithm, and the wedge trust region method is projected for derivative-free problems. We combined the non-monotone strategy into wedge trust region methods, and the computational results showed that the two strategies have their respective advantages and disadvantages.

Keywords: Non-monotone method; Derivative free optimization, Wedge trust region