English

Robins-Monro Augmented Lagrangian Method for Stochastic Convex Optimization

Optimization and Control 2022-09-02 v2

Abstract

In this paper, we propose a Robbins-Monro augmented Lagrangian method (RMALM) to solve a class of constrained stochastic convex optimization, which can be regarded as a hybrid of the Robbins-Monro type stochastic approximation method and the augmented Lagrangian method of convex optimizations. Under mild conditions, we show that the proposed algorithm exhibits a linear convergence rate. Moreover, instead of verifying a computationally intractable stopping criteria, we show that the RMALM with the increasing subproblem iteration number has a global complexity O(1/ε1+q)\mathcal{O}(1/\varepsilon^{1+q}) for the ε\varepsilon-solution (i.e., E(xkx2)<ε\mathbb{E}\left(\|x^k-x^*\|^2\right) < \varepsilon), where qq is any positive number. Numerical results on synthetic and real data demonstrate that the proposed algorithm outperforms the existing algorithms.

Keywords

Cite

@article{arxiv.2208.14019,
  title  = {Robins-Monro Augmented Lagrangian Method for Stochastic Convex Optimization},
  author = {Rui Wang and Chao Ding},
  journal= {arXiv preprint arXiv:2208.14019},
  year   = {2022}
}
R2 v1 2026-06-25T02:04:44.566Z