English

Communication Efficient Distributed Newton Method with Fast Convergence Rates

Optimization and Control 2023-05-30 v1

Abstract

We propose a communication and computation efficient second-order method for distributed optimization. For each iteration, our method only requires O(d)\mathcal{O}(d) communication complexity, where dd is the problem dimension. We also provide theoretical analysis to show the proposed method has the similar convergence rate as the classical second-order optimization algorithms. Concretely, our method can find~(ϵ,dLϵ)\big(\epsilon, \sqrt{dL\epsilon}\,\big)-second-order stationary points for nonconvex problem by O(dLϵ3/2)\mathcal{O}\big(\sqrt{dL}\,\epsilon^{-3/2}\big) iterations, where LL is the Lipschitz constant of Hessian. Moreover, it enjoys a local superlinear convergence under the strongly-convex assumption. Experiments on both convex and nonconvex problems show that our proposed method performs significantly better than baselines.

Keywords

Cite

@article{arxiv.2305.17945,
  title  = {Communication Efficient Distributed Newton Method with Fast Convergence Rates},
  author = {Chengchang Liu and Lesi Chen and Luo Luo and John C. S. Lui},
  journal= {arXiv preprint arXiv:2305.17945},
  year   = {2023}
}

Comments

Accepted in SIGKDD 2023

R2 v1 2026-06-28T10:49:01.280Z