AISTATS 2022poster38 citations

Robust Training in High Dimensions via Block Coordinate Geometric Median Descent

Anish Acharya, Abolfazl Hashemi, Prateek Jain, Sujay Sanghavi, Inderjit S. Dhillon, Ufuk Topcu

Abstract

Geometric median (GM) is a classical method in statistics for achieving robust estimation of the uncorrupted data; under gross corruption, it achieves the optimal breakdown point of 1/2. However, its computational complexity makes it infeasible for robustifying stochastic gradient descent (SGD) in high-dimensional optimization problems. In this paper, we show that by applying GM to only a judiciously chosen block of coordinates at a time and using a memory mechanism, one can retain the breakdown point of 1/2 for smooth non-convex problems, with non-asymptotic convergence rates comparable to the SGD with GM while resulting in significant speedup in training. We further validate the run-time and robustness of our approach empirically on several popular deep learning tasks. Code available at: https://github.com/anishacharya/BGMD

BibTeX
@InProceedings{pmlr-v151-acharya22a,
  title = 	 { Robust Training in High Dimensions via Block Coordinate Geometric Median Descent },
  author =       {Acharya, Anish and Hashemi, Abolfazl and Jain, Prateek and Sanghavi, Sujay and Dhillon, Inderjit S. and Topcu, Ufuk},
  booktitle = 	 {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics},
  pages = 	 {11145--11168},
  year = 	 {2022},
  editor = 	 {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel},
  volume = 	 {151},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {28--30 Mar},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v151/acharya22a/acharya22a.pdf},
  url = 	 {https://proceedings.mlr.press/v151/acharya22a.html},
  abstract = 	 { Geometric median (GM) is a classical method in statistics for achieving robust estimation of the uncorrupted data; under gross corruption, it achieves the optimal breakdown point of 1/2. However, its computational complexity makes it infeasible for robustifying stochastic gradient descent (SGD) in high-dimensional optimization problems. In this paper, we show that by applying GM to only a judiciously chosen block of coordinates at a time and using a memory mechanism, one can retain the breakdown point of 1/2 for smooth non-convex problems, with non-asymptotic convergence rates comparable to the SGD with GM while resulting in significant speedup in training. We further validate the run-time and robustness of our approach empirically on several popular deep learning tasks. Code available at: https://github.com/anishacharya/BGMD }
}
Robust Training in High Dimensions via Block Coordinate Geometric Median Descent · AISTATS 2022