English

$\bar{G}_{mst}$:An Unbiased Stratified Statistic and a Fast Gradient Optimization Algorithm Based on It

Machine Learning 2021-10-08 v1 Machine Learning

Abstract

-The fluctuation effect of gradient expectation and variance caused by parameter update between consecutive iterations is neglected or confusing by current mainstream gradient optimization algorithms. The work in this paper remedy this issue by introducing a novel unbiased stratified statistic \ Gˉmst\bar{G}_{mst}\ , a sufficient condition of fast convergence for \ Gˉmst\bar{G}_{mst}\ also is established. A novel algorithm named MSSG designed based on \ Gˉmst\bar{G}_{mst}\ outperforms other sgd-like algorithms. Theoretical conclusions and experimental evidence strongly suggest to employ MSSG when training deep model.

Keywords

Cite

@article{arxiv.2110.03354,
  title  = {$\bar{G}_{mst}$:An Unbiased Stratified Statistic and a Fast Gradient Optimization Algorithm Based on It},
  author = {Aixiang Chen},
  journal= {arXiv preprint arXiv:2110.03354},
  year   = {2021}
}
R2 v1 2026-06-24T06:42:03.296Z