- Convergence analysis of an accelerated stochastic ADMM with larger stepsizes Jianchao Bai (bjc1987163.com) Abstract: In this paper, we develop an accelerated stochastic Alternating Direction Method of Multipliers (ADMM) for solving the structured convex optimization problem whose objective is the sum of smooth and nonsmooth terms. The proposed algorithm uses the ideas of both ADMM and accelerated stochastic gradient methods, and the involved dual variables are updated symmetrically with a larger stepsize region than the so-called $(0, \frac{\sqrt{5}+1}{2})$. To simplify computation of the nonsmooth subproblem, we add an nonnegative proximal term to transform the subproblem into a proximal mapping. By a variational analysis, we show that the proposed stochastic algorithm converges in expectation with the worst-case $\C{O}(1/T)$ convergence rate, where $T$ denotes the number of outer iterations. Convergence analysis of a special stochastic augmented Lagrange method and 3-block extensions of the proposed algorithm are also investigated. Numerical experiments on testing a big-data problem in machine learning verify that the proposed method is very promising. Keywords: convex optimization, stochastic ADMM, symmetric update, larger stepsize, proximal term, complexity Category 1: Convex and Nonsmooth Optimization Category 2: Global Optimization (Stochastic Approaches ) Citation: Download: [PDF]Entry Submitted: 01/01/2020Entry Accepted: 01/01/2020Entry Last Modified: 02/12/2020Modify/Update this entry Visitors Authors More about us Links Subscribe, Unsubscribe Digest Archive Search, Browse the Repository Submit Update Policies Coordinator's Board Classification Scheme Credits Give us feedback Optimization Journals, Sites, Societies Optimization Online is supported by the Mathematical Optmization Society.