I have a pretrained CNN model using the keras library. I now need to perform a Projected Gradient Descent (PGD) to develop some adversarial examples. To do this, I will need to perform a gradient descent on the model to then project it back inside a user defined constraint (the 'projected' part of PGD). Is there a way to do this (specifically the gradient descent part) on the current model I have?
Thanks in advance