一种求解约束优化问题的新的罚信赖域算法(英文)

A New Penalty Trust-Region Method for Constrained Optimization Problems

很多现实的优化问题都能描述成一个约束优化问题。目前,已有很多经典的方法用来解决这类问题,其中尤以罚函数法最受青睐。本文中,提出了一种新的罚信赖域算法。新算法构造了两个光滑函数用来近似罚函数,并结合了信赖域算法。最后也对算法收敛性做了一些讨论。

Many real-world search and optimization problems can be formulated as constrained problems. In trying to solve them, a lot of classical methods have be developed, in which penalty function methods are the most popular. This paper presents a new penalty trust region method with a smooth approximate technique for constrained problems. The approximate technique uses two new smooth functions to deal with the problem that the original functions are not differentiable at the origin point. Several convergence results are also discussed.