共轭梯度方法是求解大规模无约束非线性优化问题的一种重要方法.对参数βk不同的构造方法,形成了各种各样的共轭梯度算法.基于现有的研究结果启发,给出了参数βk的一种新的构造方法,进而提出了一种新的共轭梯度算法.该算法能够保证目标函数序列的充分下降性,并在目标函数可微的条件下,证明了算法的全局收敛性.
Conjugate gradient is an important method that is used to solve the problem of large-scale unconstrained nonlinear optimization.The structure of the parameter βk is different,which forms a kind of conjugate gradient algorithm.Based on the inspiration of existing research results,a new constructor of the parameter βk is given,then a new conjugate gradient algorithm is proposed.It is proved that the new method is of full descent under the exact line search.Meanwhile,a method of global convergence is proved under the condition that the objective function is continuously differentiable.