研究了无约束极大极小问题.通过引入一个可微的辅助函数,利用广义投影技术产生下降搜索方向,结合Armjio非精确线搜索建立了一个广义梯度投影算法.在初始点任意的条件下,证明了算法的全局收敛性.
In this work, unconstrained minimax problems are studied. By use of proposing a differentially auxiliary function and providing explicit search direction with the aid idea of generalized gradient project technique, a new algorithm with Armjio non-exact linear search is presented and its global convergence is obtained under arbitrary initial point condition.