Corpus ID: 173188378

Are Labels Required for Improving Adversarial Robustness?

@article{Uesato2019AreLR,
  title={Are Labels Required for Improving Adversarial Robustness?},
  author={Jonathan Uesato and Jean-Baptiste Alayrac and Po-Sen Huang and Robert Stanforth and Alhussein Fawzi and Pushmeet Kohli},
  journal={ArXiv},
  year={2019},
  volume={abs/1905.13725}
}
  • Jonathan Uesato, Jean-Baptiste Alayrac, +3 authors Pushmeet Kohli
  • Published 2019
  • Computer Science, Mathematics
  • ArXiv
  • Recent work has uncovered the interesting (and somewhat surprising) finding that training models to be invariant to adversarial perturbations requires substantially larger datasets than those required for standard classification. [...] Key Method On standard datasets like CIFAR-10, a simple Unsupervised Adversarial Training (UAT) approach using unlabeled data improves robust accuracy by 21.7% over using 4K supervised examples alone, and captures over 95% of the improvement from the same number of labeled…Expand Abstract

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 60 CITATIONS

    Adversarially Robust Generalization Just Requires More Unlabeled Data

    VIEW 4 EXCERPTS
    CITES BACKGROUND & RESULTS
    HIGHLY INFLUENCED

    Unlabeled Data Improves Adversarial Robustness

    VIEW 9 EXCERPTS
    CITES METHODS & BACKGROUND
    HIGHLY INFLUENCED

    Overfitting in adversarially robust deep learning

    VIEW 2 EXCERPTS
    CITES METHODS

    Adversarial Distributional Training for Robust Deep Learning

    VIEW 1 EXCERPT
    CITES BACKGROUND

    FILTER CITATIONS BY YEAR

    2019
    2020

    CITATION STATISTICS

    • 5 Highly Influenced Citations

    • Averaged 30 Citations per year from 2019 through 2020

    • 61% Increase in citations per year in 2020 over 2019

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 57 REFERENCES

    Unlabeled Data Improves Adversarial Robustness

    VIEW 6 EXCERPTS
    HIGHLY INFLUENTIAL

    Adversarial Machine Learning at Scale

    VIEW 10 EXCERPTS
    HIGHLY INFLUENTIAL

    Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning