Title :
Convergence of modified conjugate gradient methods without line search
Author_Institution :
Feixian Sch., Linyi Normal Univ., Linyi, China
Abstract :
In this paper, a class of modified conjugate gradient methods are proposed, which have the following attractive properties: (1) the step length is determined by a formula; (2) the iterative direction is always a sufficient descent direction without utilizing the line search. Under the boundedness of the level set and the Lipschitz continuity of the underlying function, the proposed methods are global convergent. Some numerical results are given to illustrate the effectiveness of the proposed methods.
Keywords :
conjugate gradient methods; convergence of numerical methods; optimisation; set theory; Lipschitz function continuity; conjugate gradient methods; convergence; iterative direction; level set; line search; step length;
Conference_Titel :
Computational Intelligence and Natural Computing Proceedings (CINC), 2010 Second International Conference on
Conference_Location :
Wuhan
Print_ISBN :
978-1-4244-7705-0
DOI :
10.1109/CINC.2010.5643723