PGD adversarial training

[fatimah-aloraini]

Thank you very much for ART; it was incredibly helpful for my project. However, I have a couple of concerns:

1. Do you have any notebook examples for PGD adversarial training?

2. From the code, it appears that PGD adversarial training trains the model only on adversarial samples. Could you please clarify how clean samples are handled?

3. I noticed that all the adversarial training methods deal with differential models. What about other models, such as Decision Trees and Random Forests ? Is there any way to train them adversarially using ART?

I greatly appreciate your effort. Thank you.
Fatimah