非线性优化一个超线性收敛的广义投影型可行方向法

A Generalized Projection Feasible Method with Superlinear Convergence for Nonlinear Optimization

本文针对非线性不等式约束优化问题,提出了一个新的广义投影型可行方向法.该算法构造了一个新的广义投影方向,无需计算有效积极约束集,因而其计算工作量较小.在适当的条件下,证明了算法具有全局收敛及超线性收敛性.

In this paper, a new generalized projection feasible algorithm is proposed for solving the nonlinear inequality constrained optimization. A new projection direction is proposed. We don＇t compute an c active set, so the computationM cost is reduced. The theoretical analysis shows that the algorithm is globM and superlinear convergence under some suitable conditions.