【摘要】 不带线搜索的共轭梯度方法即给出步长的具体公式来代替线搜索,由Sun和Zhang[1]首次提出。 Sun 和Cheng[2]证明了不带线搜索的两参数簇共轭梯度方法的全局收敛性。本工作深入了他们的研究;还借用拟牛顿方法的思想,改进了不带线搜索的两参数簇共轭梯度方法,并给出了具体算法和数值结果。从数值结果可以看出,改进的不带线搜索的两参数簇共轭梯度方法是很有效的。更多还原

【Abstract】 The no-line-search skill for step length applied in conjugate gradient methods was first given by Sun and Zhang [1] . Cheng and Sun [2] gave the convergence property of two-parameter family of CG methods. This work, under their results, the two-parameter family of CG methods by Quasi Newton Method had been improved. In addition, the algorithm and computational results were also presented. Computational results showed that the modified two-parameter family of CG methods without line search can be more effective.更多还原