基于近似一阶信息的加速的bundle level算法
发布时间:2018-04-22 06:19
本文选题:加速算法 + bundle ; 参考:《中国科学:数学》2017年10期
【摘要】:本文提出了四种加速的BL(bundle level)算法来分别求解凸光滑函数、强凸光滑函数的极小值问题和一类鞍点(saddle-point)问题.这些算法可以运用目标函数的近似的一阶信息来得到上述几类问题的近似解.本文重点研究了在一阶信息误差上界可自由选取和给定不变的两种情形下,所提出的算法中近似解能达到的最佳精度以及相应的迭代复杂度.
[Abstract]:In this paper, we propose four accelerated BL(bundle level algorithms for solving convex smooth functions, strongly convex smooth functions and saddle-point saddle-point problems, respectively. These algorithms can use the first order information of the objective function to obtain the approximate solutions of the above problems. In this paper, we focus on the optimal accuracy of the approximate solution and the corresponding iterative complexity in the case that the upper bound of the first-order information error can be freely selected and given invariant.
【作者单位】: Department
【基金】:美国国家科学基金(批准号:DMS-1319050和DMS-1719932)资助项目
【分类号】:O174.13
,
本文编号:1785979
本文链接:https://www.wllwen.com/kejilunwen/yysx/1785979.html