On Zeroth-Order Stochastic Convex Optimization via Random Walks
Machine Learning
2014-02-13 v1 Machine Learning
Abstract
We propose a method for zeroth order stochastic convex optimization that attains the suboptimality rate of after queries for a convex bounded function . The method is based on a random walk (the \emph{Ball Walk}) on the epigraph of the function. The randomized approach circumvents the problem of gradient estimation, and appears to be less sensitive to noisy function evaluations compared to noiseless zeroth order methods.
Cite
@article{arxiv.1402.2667,
title = {On Zeroth-Order Stochastic Convex Optimization via Random Walks},
author = {Tengyuan Liang and Hariharan Narayanan and Alexander Rakhlin},
journal= {arXiv preprint arXiv:1402.2667},
year = {2014}
}
Comments
10 pages, 3 figures