非强凸问题嵌套式分散梯度方法的收敛结果
最优化与控制
2022-06-28 v2
摘要
我们关注为解决分布式优化问题而引入的NEAR-DGD(嵌套精确交替递归分布式梯度下降)方法的收敛性。在局部目标函数强凸且其梯度Lipschitz连续的假设下,\cite{BBKW - Near DGD}中建立了线性收敛。本文中,我们研究NEAR-DGD在缺乏强凸性时的收敛性质。更确切地说,我们在以下两种情形下建立收敛结果:(1)仅假设目标函数的凸性时。(2)当目标函数表示为强凸函数与秩亏矩阵的合成函数时,其属于凸且拟强凸函数类。我们提供了数值结果以支持这些收敛结论。
引用
@article{arxiv.2108.02129,
title = {Convergence results of a nested decentralized gradient method for non-strongly convex problems},
author = {Woocheol Choi and Doheon Kim and Seok-Bae Yun},
journal= {arXiv preprint arXiv:2108.02129},
year = {2022}
}
备注
28 pages, Typos were fixed and new convergence results were added for the case when the communication number is constant, to appear in J. Optim. Theory Appl