摘要
该文结合文献[1]Chen和Fukushima的邻近点拟牛顿方法和过滤集技术,给出了一个求解不可微凸优化问题的新算法.与Chen和Fukushima的方法不同,新算法不用线搜索,而是用过滤集构造接受准则,并借助于过滤集技术,证明了算法的整体收敛性.
A new algorithm for nondifferentiable convex optimization, combining the proximal quasi-Newton method with filter methods, is presented in this paper. The new algorithm uses the filter instead of using the line searching method to accept the trial step, so it is different from proximal quasi-Newton method. The global convergence of the new algorithm has been proved in this study.
出处
《南京晓庄学院学报》
2012年第6期8-12,共5页
Journal of Nanjing Xiaozhuang University
关键词
非光滑凸优化
邻近点方法
拟牛顿方法
过滤集
捆集
nondifferentiable convex optimization
proximal point method
quasi-Newton method
filter
bundle