English

Stochastic Halpern Iteration with Variance Reduction for Stochastic Monotone Inclusions

Optimization and Control 2023-01-10 v4 Machine Learning

Abstract

We study stochastic monotone inclusion problems, which widely appear in machine learning applications, including robust regression and adversarial learning. We propose novel variants of stochastic Halpern iteration with recursive variance reduction. In the cocoercive -- and more generally Lipschitz-monotone -- setup, our algorithm attains ϵ\epsilon norm of the operator with O(1ϵ3)\mathcal{O}(\frac{1}{\epsilon^3}) stochastic operator evaluations, which significantly improves over state of the art O(1ϵ4)\mathcal{O}(\frac{1}{\epsilon^4}) stochastic operator evaluations required for existing monotone inclusion solvers applied to the same problem classes. We further show how to couple one of the proposed variants of stochastic Halpern iteration with a scheduled restart scheme to solve stochastic monotone inclusion problems with O(log(1/ϵ)ϵ2){\mathcal{O}}(\frac{\log(1/\epsilon)}{\epsilon^2}) stochastic operator evaluations under additional sharpness or strong monotonicity assumptions.

Keywords

Cite

@article{arxiv.2203.09436,
  title  = {Stochastic Halpern Iteration with Variance Reduction for Stochastic Monotone Inclusions},
  author = {Xufeng Cai and Chaobing Song and Cristóbal Guzmán and Jelena Diakonikolas},
  journal= {arXiv preprint arXiv:2203.09436},
  year   = {2023}
}
R2 v1 2026-06-24T10:17:21.473Z