非光滑单值优化的非单调线搜索方法

摘要

We present a nonmonotone line search algorithm for nonsmooth unary optimization problems.rnBased on the duality theorem of linear programming, the directional derivatives of the objective function rncan be expressed as a linear programming which is very important in the practical calculation for nonmonorntone line search subproblems. A theoretical analysis proves that the proposed algorithm is globally converrngent and has a local superlinear rate under some mild conditions.%提供了非光滑单值优化的非单调线搜索方法.基于线性规划的对偶定理,目标函数的方向导数可以表示成线性规划问题,这在实际计算非单调线搜索子问题时是非常重要的.在合理的条件下,证明了算法的整体收敛性和局部超线性收敛速率.