在共轭梯度法的实际使用中, 通常在迭代 ?? 步或 ?? + 1 步之后, 重新取负梯
度方向作为搜索方向, 我们称之为再开始共轭梯度法. 这是因为对于一般非二次
函数而言, ?? 步迭代后共轭梯度法产生的搜索方向往往不再具有共轭性. 而对于
大规模问题, 常常每 ??(?? < ?? 或 ?? ? ??) 步就进行再开始. 此外, 当搜索方向不
是下降方向时, 也插入负梯度方向作为搜索方向.(In the practical use of conjugate gradient method, the negative ladder is usually taken after the iteration step or + 1 step
The degree direction is used as the search direction, and we call it the re starting conjugate gradient method. This is because of the general non two times
In terms of function, the search direction produced by conjugate gradient method is no longer conjugate after step iteration
Large scale problems, often each (< or the) step re start. In addition, when the search direction
The direction of descent is also inserted into the negative gradient direction as the search direction)