摘要
根据一类基于新拟牛顿方程Bk+1sk=yk*的修改BFGS类算法,采用广义W olfe线搜索模型(GW搜索模型):f(xk+1)≤f(xk)+δkαgTkdk和g(xk+1)Tdk≥m ax{,σ1-(kα‖dk‖)p}gTkdk,其中0<δ≤σ<1,p∈(-∞,1),得到一类修正的BFGS算法(M BFGS),证明了M BFGS算法的全局收敛性和超线性收敛性.数值试验结果表明M BFGS算法是有效的.
In this paper, we present a modified BFGS method, which satisfies the quasi-Newton funtion proposed by Wei. Under suitable conditions, we establish global convergence and superlinear convergence for our algorithm with the general Wolfe line search. The numerical results are also presented, which show that the proposed algorithm is efficient for unconstrained optimization problems.
出处
《广西科学》
CAS
2006年第4期282-287,292,共7页
Guangxi Sciences
基金
国家自然科学基金(No.10161002)
广西自然科学基金项目(No.0135004)资助
关键词
无约束优化
BFGS算法
全局收敛性
超线性收敛性
unconstrained optimization,BFGS method,global convergence, superlinear convergence
