This is the repo for attack randomized models
- CIFAR10
- VGG16
Plain
: No defense - VGG16
RSE
: Random Self-ensemble - VGG16
Adv
: Adversarial training - VGG16
Adv_vi
: Adversarial training Bayesian neural network
python main.py --attack "NES" --model "rse" --steps 100 --eps 0.2
python main.py --help