Optimal Learning for Multi-Pass Stochastic Gradient Methods

Abstract

We analyze the learning properties of the stochastic gradient method when multiple passes over the data and mini-batches are allowed. In particular, we consider the square loss and show that for a universal step-size choice, the number of passes acts as a regularization parameter, and optimal finite sample bounds can be achieved by early-stopping. Moreover, we show that larger step-sizes are allowed when considering mini-batches. Our analysis is based on a unifying approach, encompassing both batch and stochastic gradient methods as special cases.

Cite

Text

Lin and Rosasco. "Optimal Learning for Multi-Pass Stochastic Gradient Methods." Neural Information Processing Systems, 2016.

Markdown

[Lin and Rosasco. "Optimal Learning for Multi-Pass Stochastic Gradient Methods." Neural Information Processing Systems, 2016.](https://mlanthology.org/neurips/2016/lin2016neurips-optimal/)

BibTeX

@inproceedings{lin2016neurips-optimal,
  title     = {{Optimal Learning for Multi-Pass Stochastic Gradient Methods}},
  author    = {Lin, Junhong and Rosasco, Lorenzo},
  booktitle = {Neural Information Processing Systems},
  year      = {2016},
  pages     = {4556-4564},
  url       = {https://mlanthology.org/neurips/2016/lin2016neurips-optimal/}
}