English

Efficient learning with robust gradient descent

Machine Learning 2018-10-16 v3

Abstract

Minimizing the empirical risk is a popular training strategy, but for learning tasks where the data may be noisy or heavy-tailed, one may require many observations in order to generalize well. To achieve better performance under less stringent requirements, we introduce a procedure which constructs a robust approximation of the risk gradient for use in an iterative learning routine. Using high-probability bounds on the excess risk of this algorithm, we show that our update does not deviate far from the ideal gradient-based update. Empirical tests using both controlled simulations and real-world benchmark data show that in diverse settings, the proposed procedure can learn more efficiently, using less resources (iterations and observations) while generalizing better.

Keywords

Cite

@article{arxiv.1706.00182,
  title  = {Efficient learning with robust gradient descent},
  author = {Matthew J. Holland and Kazushi Ikeda},
  journal= {arXiv preprint arXiv:1706.00182},
  year   = {2018}
}