一种近似BFGS的自适应双参数共轭梯度法

An Adaptive Two-Parameter Conjugate Gradient Method for Approximating BFGS

为了更加有效的求解大规模无约束优化问题,本文基于自调比无记忆BFGS拟牛顿法,提出一个自适应双参数共轭梯度法,设计的搜索方向满足充分下降性,在一般假设和标准Wolfe线搜索准则下,证明该方法具有全局收敛性,数值实验结果证明提出的新算法是有效的.

In order to solve large-scale unconstrained optimization problems more effectively,this paper proposes an adaptive two-parameter conjugate gradient method based on the self-scaling memoryless BFGS quasi-Newton method,and the designed search direction satisfies sufficient descending,and under the general assumption and standard Wolfe line search criterion,it is proved that the proposed method has global convergence,and numerical experimental results prove that the proposed new algorithm is effective.