English

Optimal Gradient Sliding and its Application to Distributed Optimization Under Similarity

Optimization and Control 2022-05-31 v1 Distributed, Parallel, and Cluster Computing Machine Learning

Abstract

We study structured convex optimization problems, with additive objective r:=p+qr:=p + q, where rr is (μ\mu-strongly) convex, qq is LqL_q-smooth and convex, and pp is LpL_p-smooth, possibly nonconvex. For such a class of problems, we proposed an inexact accelerated gradient sliding method that can skip the gradient computation for one of these components while still achieving optimal complexity of gradient calls of pp and qq, that is, O(Lp/μ)\mathcal{O}(\sqrt{L_p/\mu}) and O(Lq/μ)\mathcal{O}(\sqrt{L_q/\mu}), respectively. This result is much sharper than the classic black-box complexity O((Lp+Lq)/μ)\mathcal{O}(\sqrt{(L_p+L_q)/\mu}), especially when the difference between LqL_q and LqL_q is large. We then apply the proposed method to solve distributed optimization problems over master-worker architectures, under agents' function similarity, due to statistical data similarity or otherwise. The distributed algorithm achieves for the first time lower complexity bounds on {\it both} communication and local gradient calls, with the former having being a long-standing open problem. Finally the method is extended to distributed saddle-problems (under function similarity) by means of solving a class of variational inequalities, achieving lower communication and computation complexity bounds.

Keywords

Cite

@article{arxiv.2205.15136,
  title  = {Optimal Gradient Sliding and its Application to Distributed Optimization Under Similarity},
  author = {Dmitry Kovalev and Aleksandr Beznosikov and Ekaterina Borodich and Alexander Gasnikov and Gesualdo Scutari},
  journal= {arXiv preprint arXiv:2205.15136},
  year   = {2022}
}

Comments

24 pages, 2 new algorithms, 12 theorems, 2 figures

R2 v1 2026-06-24T11:33:12.190Z