加速原始-对偶方法的统一变分视角
Unifying Variational View of Accelerated Primal-Dual Methods
浏览论文内容
中文总结 AI 辅助
本文提出统一变分框架,通过ELR原理生成加速原始-对偶流,涵盖镜像与ADMM特例,并证明指数收敛,扩展至概率分布优化。
中文摘要 AI 辅助
我们为仿射约束凸优化中的加速原始-对偶流开发了一个统一的变分框架。特别地,我们证明了将欧拉-拉格朗日-瑞利(ELR)原理应用于耦合增广Bregman拉格朗日函数,能够系统地生成一般Bregman几何上的加速动力学,并恢复一系列现有的原始-对偶镜像流和交替方向乘子法(ADMM)流作为特例。随后,我们在温和假设下为这些流建立了$\mathcal{O}(e^{-b_t})$收敛保证。最后,我们展示了所提出的算法开发框架从有限维优化扩展到概率分布上的约束优化。
英文摘要
We develop a unifying variational framework for accelerated primal-dual flows in affinely constrained convex optimization. In particular, it is shown that applying the Euler-Lagrange-Rayleigh (ELR) principle to coupled augmented Bregman Lagrangians systematically generates accelerated dynamics over general Bregman geometries and recovers a family of existing primal-dual mirror and Alternating Direction Method of Multipliers (ADMM) flows as special cases. We then establish $\mathcal{O}(e^{-b_t})$ convergence guarantees for these flows under mild assumptions. Lastly, we show that the proposed framework for algorithmic development extends from finite-dimensional optimization to constrained optimization over probability distributions.
发表机构
- William E. Boeing Department of Aeronautics and Astronautics, University of Washington(华盛顿大学威廉E.波音航空航天系)
机构由 AI 辅助整理,请以论文原文为准。