摘要
本文给出了凸二次优化问题基于一类有限核函数的新的大步校正内点算法.这些核函数是一类相当广泛的函数,它的主要特征是非自正则的,而且在其可行域边界上的值是有限的.利用类似于线性规划的相应算法的分析方法,证明了新算法具有目前最好的大步校正算法的迭代复杂性,即O(nlognlog(n/ε)).
A new large-update interior-point algorithm for convex quadratic optimization based on the class of finite kernel functions is presented.These proposed functions are a kind of universal functions.Their main features are not self-regular,and that the value at the boundary of their feasible region is finite.By using the analysis which is analogous to the corresponding method for linear optimization;the iteration complexity of the new algorithm is shown to be O（√nlognlog（n/ε））,which is currently the best known complexity for large-update methods.
出处
《三峡大学学报（自然科学版）》
CAS
2009年第6期104-109,共6页
Journal of China Three Gorges University：Natural Sciences
基金
湖北省自然科学基金项目(2008CDZ047)
关键词
凸二次优化
核函数
内点算法
大步校正算法
多项式复杂性
convex quadratic optimization
kernel function
interior-point method
large-update method
polynomial complexity
