›› 2013, Vol. 26 ›› Issue (11): 4-.

• Articles • Previous Articles     Next Articles

A Quasi-Newton Nonmonotone Trust Region Method of New Conic Model

LI Xiaowei,QIAN Huimin   

  1. (School of Science,Xidian University,Xi'an 710071,China)
  • Online:2013-11-15 Published:2013-11-19

Abstract:

A trust region algorithm based on new conic model is presented for unconstrained optimization by combining the nonmonotonic wolfe line search and quasi-newton technique.A new trust region subproblem is constructed,in which the trust region radius uses the information of gk and Bk.The subproblem is solved by using cp||B-1k+1|| ||gk+1||],c∈(0,1),p is a nonnegative integer.Therefore,instead of adjusting Δk,one adjusts p for each iteration.Under proper assumptions,the global convergence of the method is proved.

Key words: unconstrained optimization;nonmonotonic line search;quasi-newton;new conic model;trust region method;global convergence

CLC Number: 

  • O221