A two-layer network trained on mixed clean and perturbed logits recovers original predictions for a range of adversarial attacks without needing image data.
Training deep neural-networks using a noise adaptation layer
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Defending Adversarial Attacks by Correcting logits
A two-layer network trained on mixed clean and perturbed logits recovers original predictions for a range of adversarial attacks without needing image data.