Exponential convergence of testing error for stochastic gradient methods
Machine Learning
2018-11-21 v4 Machine Learning
Abstract
We consider binary classification problems with positive definite kernels and square loss, and study the convergence rates of stochastic gradient methods. We show that while the excess testing loss (squared loss) converges slowly to zero as the number of observations (and thus iterations) goes to infinity, the testing error (classification error) converges exponentially fast if low-noise conditions are assumed.
Cite
@article{arxiv.1712.04755,
title = {Exponential convergence of testing error for stochastic gradient methods},
author = {Loucas Pillaud-Vivien and Alessandro Rudi and Francis Bach},
journal= {arXiv preprint arXiv:1712.04755},
year = {2018}
}