A revision of the rectangular algorithm for a class of DC optimization problems
A revision of the rectangular algorithm for a class of DC optimization problems
复制标题
一类DC优化问题的矩形算法的修正
DOI：
10.1007/s10898-021-01102-2
复制
发表时间：
2021
影响因子：
1.8
通讯作者：
Kuno Takahito
中科院分区：
文献类型：
--
作者：
Homma Midori;Miura Misa;Hirayama Yo;Takahashi Tamao;Miura Takahiro;Yoshida Naoki;Miyata Satoshi;Kohzuki Masahiro;Ebihara Satoru;矢田和善・石井晶・青嶋誠;呉偉;Kuno Takahito
每一个连续可微函数都可以表示为一个凸函数和一个可加可分凸函数之差。我们证明了具有这种结构的DC函数可以用可分非凸优化的矩形算法来优化，并为实际应用发展了该算法的修正。我们还报告了一些数值结果，这些结果表明了修正的有效性。
Every continuously differentiable function can be represented as a difference between a convex function and an additively separable convex function. We show that a DC function with this structure can be optimized using the rectangular algorithm for separable nonconvex optimization, and develop a revision to this algorithm for practical use. We also report some numerical results which indicate the effectiveness of the revision.