arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.04161math.OC
收藏

无需重启的(加速)梯度滑动方法用于强凸复合优化

Restart-Free Oracle-Efficient Methods for Strongly Convex Composite Optimization

Department of Mathematics, College of Sciences, Shanghai University(上海大学理学院数学系)

Xinming Wu, Zi Xu, Huiling Zhang

AI总结:

本文提出无需重启的加速梯度滑动算法,用于解决强凸复合优化问题,通过优化参数选择策略,实现更高效的收敛性能。

摘要:

在本文中,我们研究了一类复合优化问题,其目标函数由一个一般的光滑和非光滑组件以及一个相对简单的非光滑项之和构成。尽管在一阶方法中通常使用重启策略以在强凸性下实现最优收敛,但这些策略引入了结构复杂性和实际开销,使得算法设计和嵌套变得繁琐。为了解决这个问题,我们提出了一种无需重启的随机梯度滑动算法,当简单的非光滑组件是强凸时,该算法消除了显式重启阶段的需要。通过一种新颖且精心设计的参数选择策略,我们证明所提出的算法仅需$\mathcal{O}(\log(\frac{1}ε))$次光滑组件的梯度评估即可得到一个$ε$-解,同时对于非光滑组件,需要$\mathcal{O}(\frac{1}ε)$次随机子梯度评估,这与现有多阶段(基于重启)方法的最优复杂度相匹配。此外,对于非光滑组件结构化的情况,允许将整体问题重新表述为双线性鞍点问题,我们开发了一种无需重启的加速随机梯度滑动算法。我们证明所得到的方法仅需$\mathcal{O}(\log(\frac{1}ε))$次光滑组件的梯度计算,同时保持了解决相应鞍点问题的整体迭代复杂度为$\mathcal{O}(\frac{1}{\sqrtε})$。我们的工作因此提供了更简单、无需重启的解决方案,从而在强凸复合优化中实现了更高效的算法设计。

英文摘要:

We consider strongly convex composite optimization problems whose objective is the sum of a general smooth component, a general nonsmooth component, and a relatively simple nonsmooth term. Although existing first-order methods can achieve optimal convergence guarantees under strong convexity, they typically rely on explicit restart schemes, leading to multiphase algorithms with increased structural and implementation complexity. To address this issue, we develop a restart-free stochastic gradient sliding method that incorporates strong convexity directly through a novel parameter selection strategy. We establish that the proposed method computes an $ε$-solution with $\mathcal{O}\bigl(\log(1/ε)\bigr)$ gradient evaluations of the smooth component and $\mathcal{O}(1/ε)$ stochastic subgradient evaluations of the general nonsmooth component. These bounds match the corresponding complexity guarantees of existing multiphase, restart-based methods, while eliminating the need for explicit restarts. Furthermore, when subgradients of the general nonsmooth component are difficult to compute, but a compatible smooth approximation and its first-order oracle are available, we develop a restart-free accelerated stochastic smoothed gradient sliding (RF-ASSGS) method. RF-ASSGS requires $\mathcal{O}\bigl(\log(1/ε)\bigr)$ gradient evaluations of the smooth component and $\mathcal{O}\bigl(1/\sqrtε+σ^2/ε\bigr)$ stochastic-oracle evaluations. Overall, our results demonstrate that restart-free gradient sliding methods can preserve the optimal convergence guarantees of their restart-based counterparts for strongly convex objectives, while substantially simplifying the algorithmic structure and implementation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑