Convergence and Complexity of Stochastic Subgradient Methods with Dependent Data for Nonconvex Optimization
Convergence and Complexity of Stochastic Subgradient Methods with Dependent Data for Nonconvex Optimization
复制标题
非凸优化的具有相关数据的随机次梯度方法的收敛性和复杂性
DOI:
10.48550/arxiv.2203.15797
复制
发表时间:
2022
期刊:
影响因子:
--
通讯作者:
Hanbaek Lyu
中科院分区:
文献类型:
--
作者:
Ahmet Alacaoglu;Hanbaek Lyu
A BSTRACT . We show that under a general dependent data sampling scheme, the classical stochastic projected and proximal subgradient methods for weakly convex functions have worst-case rate of convergence ˜ O ( n − 1/4 ) and complexity ˜ O ( ε − 4 ) for achieving an ε -near stationary point in terms of the norm of the gradient of Moreau envelope. While classical convergence guarantee requires i
登录
查看更多内容
DOI:
--
发表时间:
2019-11
期刊:
ArXiv
影响因子:
--
作者:
Hanbaek Lyu;D. Needell;L. Balzano
通讯作者:
Hanbaek Lyu;D. Needell;L. Balzano
影响因子:
2.2
作者:
Tao Sun;Yuejiao Sun;Yangyang Xu;W. Yin
通讯作者:
Tao Sun;Yuejiao Sun;Yangyang Xu;W. Yin
DOI:
10.1287/moor.2017.0889
发表时间:
2016-02
期刊:
Math. Oper. Res.
影响因子:
--
作者:
D. Drusvyatskiy;A. Lewis
通讯作者:
D. Drusvyatskiy;A. Lewis
影响因子:
3
作者:
Davis, Damek;Drusvyatskiy, Dmitriy;Lee, Jason D.
通讯作者:
Lee, Jason D.
影响因子:
3.5
作者:
Baars, BJ
通讯作者:
Baars, BJ