共轭梯度法是优化方法中最常用的方法之一,适于解决大规模问题,因此有着广泛的应用。针对无约束优化问题,基于搜索方向的选择,提出了一个新的共轭梯度法,该算法在每一次迭代过程中,均可保证搜索方向的充分下降性,并在弱的wolfe条件下,证明了算法的全局收敛性,数值结果表明了算法的可行性与有效性。
The conjugate gradient method is an important method for optimization problems. It is suitable for large scale optimization problems and wide application. A new conjugate gradient method is presented based on the modified search direction. In each iteration, the sufficient descent propeity of search direction is assured. The global convergence of this method with the weak Wolfe line search rule and numerical results are reported.