Noise Augmentation Is All You Need For FGSM Fast Adversarial Training: Catastrophic Overfitting And Robust Overfitting Require Different Augmentation.
Chaoning ZhangKang ZhangAxi NiuChenshuang ZhangJiu FengChang D. YooIn So KweonPublished in: CoRR (2022)
Keyphrases
- image noise
- avoid overfitting
- training error
- decision trees
- cross validation
- noisy environments
- training process
- data sets
- noise level
- signal to noise ratio
- noisy data
- missing data
- computationally efficient
- error rate
- noise model
- training samples
- registration errors
- geometric distortions
- image watermarking
- supervised learning
- unseen data
- noise sensitivity
- support vector machine
- imaging artifacts
- salt pepper