Viet-Anh on Software Logo

What is: Adversarial Model Perturbation?

SourceRegularizing Neural Networks via Adversarial Model Perturbation
Year2000
Data SourceCC BY-SA - https://paperswithcode.com

Based on the understanding that the flat local minima of the empirical risk cause the model to generalize better. Adversarial Model Perturbation (AMP) improves generalization via minimizing the AMP loss, which is obtained from the empirical risk by applying the worst norm-bounded perturbation on each point in the parameter space.