摘要:共轭梯度法是求解大规模无约束非线性优化问题的一种重要方法。基于现有的研究结果启发,本文对搜索方向中的参数进行了修正,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,并在强Wolfe线性搜索下,证明了算法的全局收敛性。
关键词:无约束最优化;共轭梯度法;充分下降性;线性搜索;全局收敛性
ABSTRACT:Conjugate gradient method is an important method for solving large-scale unconstrained nonlinear optimization problems. Based on the existing research, this paper modifies the parameter in the research direction, and a new conjugate gradient method is proposed. The presented method has sufficiently descent property. At the end of this paper, we show the global convergence of the proposed method with strong Wolfe line search.
Keywords: unconstrained optimization; conjugate gradient method; sufficiently descent property; line search; global convergence
本文主要讨论了无约束优化问题的算法研究,介绍了几种共轭梯度经典算法,在此基础上,给出了新的参数,即构造了新的共轭下降方向,从而得到一类新的共轭梯度算法,并证明了算法的充分下降性和全局收敛性。
由于是数学论文,简介里有很多公式复制不出来。Wrod里是有公式的请放心。