Stochastic Modified Equations and Adaptive Stochastic Gradient Algorithms
Abstract
We develop the method of stochastic modified equations (SME), in which stochastic gradient algorithms are approximated in the weak sense by continuous-time stochastic differential equations. We exploit the continuous formulation together with optimal control theory to derive novel adaptive hyper-parameter adjustment policies. Our algorithms have competitive performance with the added benefit of being robust to varying models and datasets. This provides a general methodology for the analysis and design of stochastic gradient algorithms.
Cite
Text
Li et al. "Stochastic Modified Equations and Adaptive Stochastic Gradient Algorithms." International Conference on Machine Learning, 2017.Markdown
[Li et al. "Stochastic Modified Equations and Adaptive Stochastic Gradient Algorithms." International Conference on Machine Learning, 2017.](https://mlanthology.org/icml/2017/li2017icml-stochastic/)BibTeX
@inproceedings{li2017icml-stochastic,
title = {{Stochastic Modified Equations and Adaptive Stochastic Gradient Algorithms}},
author = {Li, Qianxiao and Tai, Cheng and E, Weinan},
booktitle = {International Conference on Machine Learning},
year = {2017},
pages = {2101-2110},
volume = {70},
url = {https://mlanthology.org/icml/2017/li2017icml-stochastic/}
}