English

NESTT: A Nonconvex Primal-Dual Splitting Method for Distributed and Stochastic Optimization

Optimization and Control 2017-06-06 v2 Machine Learning Machine Learning

Abstract

We study a stochastic and distributed algorithm for nonconvex problems whose objective consists of a sum of NN nonconvex Li/NL_i/N-smooth functions, plus a nonsmooth regularizer. The proposed NonconvEx primal-dual SpliTTing (NESTT) algorithm splits the problem into NN subproblems, and utilizes an augmented Lagrangian based primal-dual scheme to solve it in a distributed and stochastic manner. With a special non-uniform sampling, a version of NESTT achieves ϵ\epsilon-stationary solution using O((i=1NLi/N)2/ϵ)\mathcal{O}((\sum_{i=1}^N\sqrt{L_i/N})^2/\epsilon) gradient evaluations, which can be up to O(N)\mathcal{O}(N) times better than the (proximal) gradient descent methods. It also achieves Q-linear convergence rate for nonconvex 1\ell_1 penalized quadratic problems with polyhedral constraints. Further, we reveal a fundamental connection between primal-dual based methods and a few primal only methods such as IAG/SAG/SAGA.

Keywords

Cite

@article{arxiv.1605.07747,
  title  = {NESTT: A Nonconvex Primal-Dual Splitting Method for Distributed and Stochastic Optimization},
  author = {Davood Hajinezhad and Mingyi Hong and Tuo Zhao and Zhaoran Wang},
  journal= {arXiv preprint arXiv:1605.07747},
  year   = {2017}
}

Comments

35 pages, 2 figures

R2 v1 2026-06-22T14:08:58.724Z