English

Robust Sampling in Deep Learning

Machine Learning 2020-06-08 v2 Machine Learning

Abstract

Deep learning requires regularization mechanisms to reduce overfitting and improve generalization. We address this problem by a new regularization method based on distributional robust optimization. The key idea is to modify the contribution from each sample for tightening the empirical risk bound. During the stochastic training, the selection of samples is done according to their accuracy in such a way that the worst performed samples are the ones that contribute the most in the optimization. We study different scenarios and show the ones where it can make the convergence faster or increase the accuracy.

Keywords

Cite

@article{arxiv.2006.02734,
  title  = {Robust Sampling in Deep Learning},
  author = {Aurora Cobo Aguilera and Antonio Artés-Rodríguez and Fernando Pérez-Cruz and Pablo Martínez Olmos},
  journal= {arXiv preprint arXiv:2006.02734},
  year   = {2020}
}

Comments

8 pages, 3 figures