针对无约束优化问题,提出一种新的混合杂交共轭梯度法,该方法在不采用Wolfe搜索的条件下,保证了算法的全局收敛性,并在每次迭代过程中,均可得到初始的自适应步长和充分下降方向.数值结果表明,该算法可行、有效.
We proposed a new hybrid conjugate gradient method.It does not need to use Wolfe line search to guarantee the global convergence.An initial self-adaptive step size and sufficient descents are obtained to the function at each iteration.Numerical results show that the method is feasible and effective.