The code is mainly based on the Adversarial ML Tutorial here: https://adversarial-ml-tutorial.org/adversarial_training/


We stored the evaluations in the results folder, run the following to generate Table 1 in the paper:
   python tables.py 2dnn 0.1 10

To start from beginning, run the following to get the results for 'net = 2dnn', 'eps = 0.1', 'alpha = 10', and 'seed = 0':
   python train.py 2dnn 0.1 10 0
   python adv_train.py 2dnn 0.1 10 0
   python eval.py 2dnn 0.1 10 0

After executing the above commands for different seeds in {0,1,2,...,9}, run the following to generate Table 1 in the paper:
   python tables.py 2dnn 0.1 10

