一种修正的HS共轭梯度法及其全局收敛性

摘要

对经典的HS共轭梯度法进行了修正,保证了搜索方向的充分下降性,这一性质在非精确线搜索和非凸函数情形下也是成立的.在适当的假设下证明了强Wolfe线搜索下算法的全局收敛性,数值实验表明算法数值效果良好.%A modification has been made to HS conjugate gradient method; the sufficient descent property of search direction is guaranteed, which has also been established under the condition of inexact line search condition and assumption that the objective function is nonconvex; under mild conditions, the global convergence property has been proved with the strong Wolfe line search; the numerical testing shows that the numerical effect of algorithm is well.