作 者: ;
机构地区: 湖南大学数学与计量经济学院
出 处: 《计算数学》 2011年第3期249-256,共8页
摘 要: 基于非单调的frame概念,提出了一个求解无约束最优化问题的直接搜索共轭梯度算法.该算法不使用充分下降条件而能够在网格(grid)之外进行搜索,这一点不同于GSS(generating setsearch)算法框架,后者为了实现网格之外的搜索必须使用充分下降条件或者移动网格(movinggrids)技术或者有理点阵(rational lattice)技术.在一定的条件下,该算法的全局收敛性也得到了证明.数值试验表明,该算法是有效的. A non-monotone frame-based direct search conjugate gradient method for unconstrained minimization is proposed. Search out of the grids are allowable while no sufficient decrease condition is used. This is different from the GSS (generating set search) method which need to employ one of the following three strategies, namely sufficient decrease, rational lattice and moving grids, to search out of the grid. Under mild conditions, the global convergence is showed. Our numerical experiences show that the proposed method is efficient.