Adversarial training reduces loss sharpness and improves cross-Monte-Carlo generalization for Higgs-jet classifiers, with projected gradient descent giving the largest gains.
In this context, N is the number of steps; Πx;ϵ is the ball- projection operator, centered around x with radius ϵ; and α is the step size
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
hep-ph 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Enhancing generalization in high energy physics using white-box adversarial attacks
Adversarial training reduces loss sharpness and improves cross-Monte-Carlo generalization for Higgs-jet classifiers, with projected gradient descent giving the largest gains.