ALOE and AOE:

Code from Chen et al. Git link: https://github.com/jfc43/robust-ood-detection
Modified /utils/pgd_attack.py according to robust overfitting pgd attack code by Rice et al. Git link: https://github.com/locuslab/robust_overfitting/blob/master/train_cifar.py

Overview of the code:

For training ALOE and AOE models you can run robust_ood_train.py from CIFAR directory using the following commands.

Train an ALOE model for CIFAR10:
python robust_ood_train.py --name Cifar10_ALOE  --adv --ood

Train an AOE model for CIFAR10:
python robust_ood_train.py --name Cifar10_AOE  --adv --ood —adv-only-in

Train an ALOE model for CIFAR100:
python robust_ood_train.py --name Cifar100_ALOE  --adv --ood --in-dataset CIFAR-100

Train an AOE model for CIFAR100:
python robust_ood_train.py --name Cifar100_AOE  --adv --ood --adv-only-in --in-dataset CIFAR-100