This reporsitory is for defense homework.
Please revise the tasks/defense_homework/defense_homework.py
file to complete the task.
You can use the following commands to evaluate the result:
python Evaluator.py
Basically, the original accuracy of the adversarial examples are around 40%
and the goal of the homework is to make the accuracy of the adv examples to 90%
on the 1000 test data points.
The defense model wil be saved under models/defense_homework-model.pth
.
Important change: For submission on Gradescope, you will have to submit both the trained model (the pth file) as well as your py code. Your submission should be a zip file called defense_homework_{ucinetid}.zip
, containing 2 files: defense_homework.py
and defense_homework-model.pth
.