A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf
.
Understanding Generalization in Adversarial Training via the Bias-Variance Decomposition
[article]
2021
arXiv
pre-print
Adversarially trained models exhibit a large generalization gap: they can interpolate the training set even for large perturbation radii, but at the cost of large test error on clean samples. To investigate this gap, we decompose the test risk into its bias and variance components and study their behavior as a function of adversarial training perturbation radii (ε). We find that the bias increases monotonically with ε and is the dominant term in the risk. Meanwhile, the variance is unimodal as
arXiv:2103.09947v2
fatcat:xa45kg3ykjgcje5qtmue6rblia