Robust Descent Using Smoothed Multiplicative Noise

Abstract

In this work, we propose a novel robust gradient descent procedure which makes use of a smoothed multiplicative noise applied directly to observations before constructing a sum of soft-truncated gradient coordinates. We show that the procedure has competitive theoretical guarantees, with the major advantage of a simple implementation that does not require an iterative sub-routine for robustification. Empirical tests reinforce the theory, showing more efficient generalization over a much wider class of data distributions.

Cite

Text

Holland. "Robust Descent Using Smoothed Multiplicative Noise." Artificial Intelligence and Statistics, 2019.

Markdown

[Holland. "Robust Descent Using Smoothed Multiplicative Noise." Artificial Intelligence and Statistics, 2019.](https://mlanthology.org/aistats/2019/holland2019aistats-robust/)

BibTeX

@inproceedings{holland2019aistats-robust,
  title     = {{Robust Descent Using Smoothed Multiplicative Noise}},
  author    = {Holland, Matthew J.},
  booktitle = {Artificial Intelligence and Statistics},
  year      = {2019},
  pages     = {703-711},
  volume    = {89},
  url       = {https://mlanthology.org/aistats/2019/holland2019aistats-robust/}
}