2 code implementations • 9 Jun 2021 • Julian Büchel, Fynn Faber, Dylan R. Muir
We present a new adversarial network optimisation algorithm that attacks network parameters during training, and promotes robust performance during inference in the face of parameter variation.