Random directions stochastic approximation with deterministic perturbations

DOI10.1109/TAC.2019.2930821MaRDI QIDQ6305201arXivFDO

Authors L. A. Prashanth, Shalabh Bhatnagar, Nirav Bhavsar, Michael C. Fu, Steven I. Marcus

Publication date 8 August 2018

Published in arXiv

Full work available at URL https://doi.org/10.1109/TAC.2019.2930821

Mathematics Subject Classification ID

Perturbations in control/observation systems (93C73) Stochastic systems in control theory (general) (93E03)

arXiv classification

cs.LG math.OC

Abstract: We introduce deterministic perturbation schemes for the recently proposed random directions stochastic approximation (RDSA) [17], and propose new first-order and second-order algorithms. In the latter case, these are the first second-order algorithms to incorporate deterministic perturbations. We show that the gradient and/or Hessian estimates in the resulting algorithms with deterministic perturbations are asymptotically unbiased, so that the algorithms are provably convergent. Furthermore, we derive convergence rates to establish the superiority of the first-order and second-order algorithms, for the special case of a convex and quadratic optimization problem, respectively. Numerical experiments are used to validate the theoretical results.

This page was built for publication: Random directions stochastic approximation with deterministic perturbations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6305201)