解半定规划的带筛子的正则化方法

A regularization method with sieve for semidefinite programming

研究了求解半定规划问题的一个带有筛子的正则化方法,该方法是基于经典的二次正则化方法,将半定规划问题转化为目标函数为凸的、可微的无约束优化问题。利用筛选信赖域方法来解这个无约束优化问题,并给出算法及其收敛性分析。

This paper studies a regularization method with sieve for semidefinite programming （SDP）, which is based on the classical quadratic regularization method, that is to transform semidefinite programming （SDP） into a convex, differentiable unconstrained objective function. A method for unconstrained optimization problems is given on the basis of trust region algorithm, and the algorithm and convergence analysis are also presented.