English

Stochastic Approximation of Smooth and Strongly Convex Functions: Beyond the $O(1/T)$ Convergence Rate

Machine Learning 2019-01-29 v1 Machine Learning

Abstract

Stochastic approximation (SA) is a classical approach for stochastic convex optimization. Previous studies have demonstrated that the convergence rate of SA can be improved by introducing either smoothness or strong convexity condition. In this paper, we make use of smoothness and strong convexity simultaneously to boost the convergence rate. Let λ\lambda be the modulus of strong convexity, κ\kappa be the condition number, FF_* be the minimal risk, and α>1\alpha>1 be some small constant. First, we demonstrate that, in expectation, an O(1/[λTα]+κF/T)O(1/[\lambda T^\alpha] + \kappa F_*/T) risk bound is attainable when T=Ω(κα)T = \Omega(\kappa^\alpha). Thus, when FF_* is small, the convergence rate could be faster than O(1/[λT])O(1/[\lambda T]) and approaches O(1/[λTα])O(1/[\lambda T^\alpha]) in the ideal case. Second, to further benefit from small risk, we show that, in expectation, an O(1/2T/κ+F)O(1/2^{T/\kappa}+F_*) risk bound is achievable. Thus, the excess risk reduces exponentially until reaching O(F)O(F_*), and if F=0F_*=0, we obtain a global linear convergence. Finally, we emphasize that our proof is constructive and each risk bound is equipped with an efficient stochastic algorithm attaining that bound.

Keywords

Cite

@article{arxiv.1901.09344,
  title  = {Stochastic Approximation of Smooth and Strongly Convex Functions: Beyond the $O(1/T)$ Convergence Rate},
  author = {Lijun Zhang and Zhi-Hua Zhou},
  journal= {arXiv preprint arXiv:1901.09344},
  year   = {2019}
}